Skip to contentSkip to conversation

Tidy up vfio_device life cycle

PATCH Superseded Kevin Tian 5 participants 26 messages Sep 9, 2022 – Sep 22, 2022

Subsystems (16)
DRM drivers KVM Intel i915 s390 +12 more subsystems

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Kevin Tian

author reviewer
#0 lore raw

v3 cover Tidy up vfio_device life cycle

The idea is to let vfio core manage the vfio_device life cycle instead of duplicating the logic cross drivers. Besides cleaner code in driver side this also allows adding struct device to vfio_device as the first step toward adding cdev uAPI in the future. Another benefit is that user can now look at sysfs to decide whether a device is bound to vfio [1], e.g.:

	/sys/devices/pci0000\:6f/0000\:6f\:01.0/vfio-dev/vfio0

Though most drivers can fit the new model naturally:

- vfio_alloc_device() to allocate and initialize vfio_device
- vfio_put_device() to release vfio_device
- dev_ops->init() for driver private initialization
- dev_ops->release() for driver private cleanup

vfio-ccw is the only exception due to a life cycle mess that its private structure mixes both parent and mdev info hence must be alloc/freed outside of the life cycle of vfio device.

Per prior discussions this won't be fixed in short term by IBM folks [2].

Instead of waiting this series introduces a few tricks to move forward:

- vfio_init_device() to initialize a pre-allocated device structure;

- require *EVERY* driver to implement @release and free vfio_device inside. Then vfio-ccw can use a completion mechanism to delay the free to css driver;

The second trick is not a real burden to other drivers because they all require a @release for private cleanup anyay. Later once the ccw mess is fixed a simple cleanup can be done by moving free from @release to vfio core.

Thanks
Kevin

[1] https://listman.redhat.com/archives/libvir-list/2022-August/233482.html
[2] https://lore.kernel.org/all/[email protected]/

v3:
- rebase to vfio-next after resolving conflicts with Yishai's series
- add missing fixes for two checkpatch errors
- fix grammar issues (Eric Auger)
- add more r-b's

v2:
- https://lore.kernel.org/lkml/[email protected]/
- rebase to 6.0-rc3
- fix build warnings (lkp)
- patch1: remove unnecessary forward reference (Jason)
- patch10: leave device_set released by vfio core (Jason)
- patch13: add Suggested-by
- patch15: add ABI file sysfs-devices-vfio-dev (Alex)
- patch15: rename 'vfio' to 'vfio_group' in procfs (Jason)

v1: https://lore.kernel.org/lkml/[email protected]/

Kevin Tian (6):
vfio: Add helpers for unifying vfio_device life cycle drm/i915/gvt: Use the new device life cycle helpers vfio/platform: Use the new device life cycle helpers vfio/amba: Use the new device life cycle helpers vfio/ccw: Use the new device life cycle helpers
vfio: Rename vfio_device_put() and vfio_device_try_get()

Yi Liu (9):
vfio/pci: Use the new device life cycle helpers vfio/mlx5: Use the new device life cycle helpers vfio/hisi_acc: Use the new device life cycle helpers vfio/mdpy: Use the new device life cycle helpers vfio/mtty: Use the new device life cycle helpers vfio/mbochs: Use the new device life cycle helpers vfio/ap: Use the new device life cycle helpers vfio/fsl-mc: Use the new device life cycle helpers
vfio: Add struct device to vfio_device

 .../ABI/testing/sysfs-devices-vfio-dev        |   8 +
 MAINTAINERS                                   |   1 +
 drivers/gpu/drm/i915/gvt/gvt.h                |   5 +-
 drivers/gpu/drm/i915/gvt/kvmgt.c              |  52 ++++--
 drivers/gpu/drm/i915/gvt/vgpu.c               |  33 ++--
 drivers/s390/cio/vfio_ccw_ops.c               |  52 +++++-
 drivers/s390/cio/vfio_ccw_private.h           |   3 +
 drivers/s390/crypto/vfio_ap_ops.c             |  50 +++---
 drivers/vfio/fsl-mc/vfio_fsl_mc.c             |  85 +++++----
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c    |  80 ++++-----
 drivers/vfio/pci/mlx5/main.c                  |  50 ++++--
 drivers/vfio/pci/vfio_pci.c                   |  20 +--
 drivers/vfio/pci/vfio_pci_core.c              |  23 ++-
 drivers/vfio/platform/vfio_amba.c             |  72 ++++++--
 drivers/vfio/platform/vfio_platform.c         |  66 +++++--
 drivers/vfio/platform/vfio_platform_common.c  |  71 +++-----
 drivers/vfio/platform/vfio_platform_private.h |  18 +-
 drivers/vfio/vfio_main.c                      | 167 +++++++++++++++---
 include/linux/vfio.h                          |  28 ++-
 include/linux/vfio_pci_core.h                 |   6 +-
 samples/vfio-mdev/mbochs.c                    |  73 +++++---
 samples/vfio-mdev/mdpy.c                      |  81 +++++----
 samples/vfio-mdev/mtty.c                      |  67 ++++---
 23 files changed, 730 insertions(+), 381 deletions(-)
 create mode 100644 Documentation/ABI/testing/sysfs-devices-vfio-dev

base-commit: f39856aacb078c1c93acef011a37121b17d54fe0

Signature · 1 line
2.21.3
author reviewer
#1 lore raw

PATCH v3 1/15 vfio: Add helpers for unifying vfio_device life cycle

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()

Drivers can register @init/@release 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.

---
 drivers/vfio/vfio_main.c | 92 ++++++++++++++++++++++++++++++++++++++++
 include/linux/vfio.h     | 25 ++++++++++-
 2 files changed, 116 insertions(+), 1 deletion(-)
drivers/vfio/vfio_main.c +92 -0
index 27d9186f35d5..adc1b697bb78 100644
@@ -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)
 {
include/linux/vfio.h +24 -1
index 0e2826559091..f67cac700e6f 100644
@@ -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);
Signature · 1 line
2.21.3
#2 lore raw

Re: [PATCH v3 01/15] vfio: Add helpers for unifying vfio_device life cycle

Hi, Kevin,

Quoted 8 lines · Kevin Tian

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,
- 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

Quoted 202 lines

Drivers can register @init/@release 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.

---
  drivers/vfio/vfio_main.c | 92 ++++++++++++++++++++++++++++++++++++++++
  include/linux/vfio.h     | 25 ++++++++++-
  2 files changed, 116 insertions(+), 1 deletion(-)
drivers/vfio/vfio_main.c +92 -0
index 27d9186f35d5..adc1b697bb78 100644
@@ -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)
  {
include/linux/vfio.h +24 -1
index 0e2826559091..f67cac700e6f 100644
@@ -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);
Signature · 1 line
"firm, enduring, strong, and long-lived"
author reviewer
#3 lore raw

RE: [PATCH v3 01/15] vfio: Add helpers for unifying vfio_device life cycle

Quoted 15 lines

From: Ethan Zhao <[email protected]>
Sent: Friday, September 9, 2022 4:24 PM

Hi, Kevin,

Quoted 8 lines · Kevin Tian

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()

Quoted 6 lines

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.

Quoted 231 lines

- 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

Quoted 37 lines

Drivers can register @init/@release 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.

Quoted 2 lines
---
  drivers/vfio/vfio_main.c | 92

++++++++++++++++++++++++++++++++++++++++

drivers/vfio/vfio_main.c void vfio_uninit_group_dev(struct vfio_device

Quoted 8 lines
  include/linux/vfio.h     | 25 ++++++++++-
  2 files changed, 116 insertions(+), 1 deletion(-)
drivers/vfio/vfio_main.c +0 -0
index 27d9186f35d5..adc1b697bb78 100644
@@ -498,6 +498,98 @@ void vfio_uninit_group_dev(struct vfio_device

*device)

Quoted 113 lines
drivers/vfio/vfio_main.c +92 -0
  }
  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)
  {
include/linux/vfio.h +2 -1
index 0e2826559091..f67cac700e6f 100644
@@ -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;

include/linux/vfio.h struct vfio_device { · line 63

Quoted 6 lines
include/linux/vfio.h +2 -0
@@ -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

include/linux/vfio.h static inline int vfio_check_feature(u32 flags,

Quoted 12 lines
include/linux/vfio.h +2 -0
   * @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,

include/linux/vfio.h static inline int vfio_check_feature(u32 flags, size_t argsz, u32 supported_ops,

Quoted 6 lines
include/linux/vfio.h +3 -0
  	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)
		\

include/linux/vfio.h static inline int vfio_check_feature(u32 flags, · line 172

Quoted 1 line
include/linux/vfio.h +1 -0
+	container_of(_vfio_alloc_device(sizeof(struct dev_struct) +
	\

include/linux/vfio.h static inline int vfio_check_feature(u32 flags, · line 173

Quoted 1 line
include/linux/vfio.h +1 -0
+					BUILD_BUG_ON_ZERO(offsetof(
		\

include/linux/vfio.h static inline int vfio_check_feature(u32 flags, · line 174

Quoted 1 line
include/linux/vfio.h +1 -0
+						struct dev_struct, member)),
	\

include/linux/vfio.h static inline int vfio_check_feature(u32 flags, · line 175

Quoted 1 line
include/linux/vfio.h +1 -0
+					dev, ops),
	\
Quoted 4 lines
include/linux/vfio.h +4 -0
+		     struct dev_struct, member)
+
+int vfio_init_device(struct vfio_device *device, struct device *dev,
+		     const struct vfio_device_ops *ops);

include/linux/vfio.h static inline int vfio_check_feature(u32 flags, size_t argsz, u32 supported_ops, · line 189

Quoted 10 lines
include/linux/vfio.h +7 -0
+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"

#4 lore raw

Re: [PATCH v3 01/15] vfio: Add helpers for unifying vfio_device life cycle

Quoted 4 lines · Kevin

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.

Right, it is quite a good/common pattern to have an allocation function return a refcount to the caller.

I don't know of any naming standard for this however.

Jason

author reviewer
#5 lore raw

PATCH v3 2/15 vfio/pci: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

Also introduce two pci core helpers as @init/@release for pci drivers:

- vfio_pci_core_init_dev()
- vfio_pci_core_release_dev()

---
 drivers/vfio/pci/vfio_pci.c      | 20 +++++++++---------
 drivers/vfio/pci/vfio_pci_core.c | 35 ++++++++++++++++++++++++++++++++
 include/linux/vfio_pci_core.h    |  2 ++
 3 files changed, 47 insertions(+), 10 deletions(-)
drivers/vfio/pci/vfio_pci.c +10 -10
index d9b5c03f8d5b..1d4919edfbde 100644
@@ -127,6 +127,8 @@ static int vfio_pci_open_device(struct vfio_device *core_vdev)
 
 static const struct vfio_device_ops vfio_pci_ops = {
 	.name		= "vfio-pci",
+	.init		= vfio_pci_core_init_dev,
+	.release	= vfio_pci_core_release_dev,
 	.open_device	= vfio_pci_open_device,
 	.close_device	= vfio_pci_core_close_device,
 	.ioctl		= vfio_pci_core_ioctl,
@@ -146,20 +148,19 @@ static int vfio_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	if (vfio_pci_is_denylisted(pdev))
 		return -EINVAL;
 
-	vdev = kzalloc(sizeof(*vdev), GFP_KERNEL);
-	if (!vdev)
-		return -ENOMEM;
-	vfio_pci_core_init_device(vdev, pdev, &vfio_pci_ops);
+	vdev = vfio_alloc_device(vfio_pci_core_device, vdev, &pdev->dev,
+				 &vfio_pci_ops);
+	if (IS_ERR(vdev))
+		return PTR_ERR(vdev);
 
 	dev_set_drvdata(&pdev->dev, vdev);
 	ret = vfio_pci_core_register_device(vdev);
 	if (ret)
-		goto out_free;
+		goto out_put_vdev;
 	return 0;
 
-out_free:
-	vfio_pci_core_uninit_device(vdev);
-	kfree(vdev);
+out_put_vdev:
+	vfio_put_device(&vdev->vdev);
 	return ret;
 }
 
@@ -168,8 +169,7 @@ static void vfio_pci_remove(struct pci_dev *pdev)
 	struct vfio_pci_core_device *vdev = dev_get_drvdata(&pdev->dev);
 
 	vfio_pci_core_unregister_device(vdev);
-	vfio_pci_core_uninit_device(vdev);
-	kfree(vdev);
+	vfio_put_device(&vdev->vdev);
 }
 
 static int vfio_pci_sriov_configure(struct pci_dev *pdev, int nr_virtfn)
drivers/vfio/pci/vfio_pci_core.c +35 -0
index 0a801aee2f2d..77d33739c6e8 100644
@@ -2078,6 +2078,41 @@ static void vfio_pci_vga_uninit(struct vfio_pci_core_device *vdev)
 					      VGA_RSRC_LEGACY_MEM);
 }
 
+int vfio_pci_core_init_dev(struct vfio_device *core_vdev)
+{
+	struct vfio_pci_core_device *vdev =
+		container_of(core_vdev, struct vfio_pci_core_device, vdev);
+
+	vdev->pdev = to_pci_dev(core_vdev->dev);
+	vdev->irq_type = VFIO_PCI_NUM_IRQS;
+	mutex_init(&vdev->igate);
+	spin_lock_init(&vdev->irqlock);
+	mutex_init(&vdev->ioeventfds_lock);
+	INIT_LIST_HEAD(&vdev->dummy_resources_list);
+	INIT_LIST_HEAD(&vdev->ioeventfds_list);
+	mutex_init(&vdev->vma_lock);
+	INIT_LIST_HEAD(&vdev->vma_list);
+	INIT_LIST_HEAD(&vdev->sriov_pfs_item);
+	init_rwsem(&vdev->memory_lock);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(vfio_pci_core_init_dev);
+
+void vfio_pci_core_release_dev(struct vfio_device *core_vdev)
+{
+	struct vfio_pci_core_device *vdev =
+		container_of(core_vdev, struct vfio_pci_core_device, vdev);
+
+	mutex_destroy(&vdev->igate);
+	mutex_destroy(&vdev->ioeventfds_lock);
+	mutex_destroy(&vdev->vma_lock);
+	kfree(vdev->region);
+	kfree(vdev->pm_save);
+	vfio_free_device(core_vdev);
+}
+EXPORT_SYMBOL_GPL(vfio_pci_core_release_dev);
+
 void vfio_pci_core_init_device(struct vfio_pci_core_device *vdev,
 			       struct pci_dev *pdev,
 			       const struct vfio_device_ops *vfio_pci_ops)
include/linux/vfio_pci_core.h +2 -0
index 089b603bcfdc..0499ea836058 100644
@@ -109,6 +109,8 @@ void vfio_pci_core_close_device(struct vfio_device *core_vdev);
 void vfio_pci_core_init_device(struct vfio_pci_core_device *vdev,
 			       struct pci_dev *pdev,
 			       const struct vfio_device_ops *vfio_pci_ops);
+int vfio_pci_core_init_dev(struct vfio_device *core_vdev);
+void vfio_pci_core_release_dev(struct vfio_device *core_vdev);
 int vfio_pci_core_register_device(struct vfio_pci_core_device *vdev);
 void vfio_pci_core_uninit_device(struct vfio_pci_core_device *vdev);
 void vfio_pci_core_unregister_device(struct vfio_pci_core_device *vdev);
Signature · 1 line
2.21.3
author reviewer
#6 lore raw

PATCH v3 3/15 vfio/mlx5: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

mlx5 has its own @init/@release for handling migration cap.

---
 drivers/vfio/pci/mlx5/main.c | 50 ++++++++++++++++++++++++++----------
 1 file changed, 36 insertions(+), 14 deletions(-)
drivers/vfio/pci/mlx5/main.c +36 -14
index 759a5f5f7b3f..fd6ccb8454a2 100644
@@ -585,8 +585,35 @@ static const struct vfio_log_ops mlx5vf_pci_log_ops = {
 	.log_read_and_clear = mlx5vf_tracker_read_and_clear,
 };
 
+static int mlx5vf_pci_init_dev(struct vfio_device *core_vdev)
+{
+	struct mlx5vf_pci_core_device *mvdev = container_of(core_vdev,
+			struct mlx5vf_pci_core_device, core_device.vdev);
+	int ret;
+
+	ret = vfio_pci_core_init_dev(core_vdev);
+	if (ret)
+		return ret;
+
+	mlx5vf_cmd_set_migratable(mvdev, &mlx5vf_pci_mig_ops,
+				  &mlx5vf_pci_log_ops);
+
+	return 0;
+}
+
+static void mlx5vf_pci_release_dev(struct vfio_device *core_vdev)
+{
+	struct mlx5vf_pci_core_device *mvdev = container_of(core_vdev,
+			struct mlx5vf_pci_core_device, core_device.vdev);
+
+	mlx5vf_cmd_remove_migratable(mvdev);
+	vfio_pci_core_release_dev(core_vdev);
+}
+
 static const struct vfio_device_ops mlx5vf_pci_ops = {
 	.name = "mlx5-vfio-pci",
+	.init = mlx5vf_pci_init_dev,
+	.release = mlx5vf_pci_release_dev,
 	.open_device = mlx5vf_pci_open_device,
 	.close_device = mlx5vf_pci_close_device,
 	.ioctl = vfio_pci_core_ioctl,
@@ -604,22 +631,19 @@ static int mlx5vf_pci_probe(struct pci_dev *pdev,
 	struct mlx5vf_pci_core_device *mvdev;
 	int ret;
 
-	mvdev = kzalloc(sizeof(*mvdev), GFP_KERNEL);
-	if (!mvdev)
-		return -ENOMEM;
-	vfio_pci_core_init_device(&mvdev->core_device, pdev, &mlx5vf_pci_ops);
-	mlx5vf_cmd_set_migratable(mvdev, &mlx5vf_pci_mig_ops,
-				  &mlx5vf_pci_log_ops);
+	mvdev = vfio_alloc_device(mlx5vf_pci_core_device, core_device.vdev,
+				  &pdev->dev, &mlx5vf_pci_ops);
+	if (IS_ERR(mvdev))
+		return PTR_ERR(mvdev);
+
 	dev_set_drvdata(&pdev->dev, &mvdev->core_device);
 	ret = vfio_pci_core_register_device(&mvdev->core_device);
 	if (ret)
-		goto out_free;
+		goto out_put_vdev;
 	return 0;
 
-out_free:
-	mlx5vf_cmd_remove_migratable(mvdev);
-	vfio_pci_core_uninit_device(&mvdev->core_device);
-	kfree(mvdev);
+out_put_vdev:
+	vfio_put_device(&mvdev->core_device.vdev);
 	return ret;
 }
 
@@ -628,9 +652,7 @@ static void mlx5vf_pci_remove(struct pci_dev *pdev)
 	struct mlx5vf_pci_core_device *mvdev = mlx5vf_drvdata(pdev);
 
 	vfio_pci_core_unregister_device(&mvdev->core_device);
-	mlx5vf_cmd_remove_migratable(mvdev);
-	vfio_pci_core_uninit_device(&mvdev->core_device);
-	kfree(mvdev);
+	vfio_put_device(&mvdev->core_device.vdev);
 }
 
 static const struct pci_device_id mlx5vf_pci_table[] = {
Signature · 1 line
2.21.3
author reviewer
#7 lore raw

PATCH v3 4/15 vfio/hisi_acc: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

Tidy up @probe so all migration specific initialization logic is moved to migration specific @init callback.

Remove vfio_pci_core_{un}init_device() given no user now.

---
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c    | 80 +++++++++----------
 drivers/vfio/pci/vfio_pci_core.c              | 30 -------
 include/linux/vfio_pci_core.h                 |  4 -
 3 files changed, 37 insertions(+), 77 deletions(-)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +37 -43
index 258cae0863ea..47174e2b61bd 100644
@@ -1213,8 +1213,28 @@ static const struct vfio_migration_ops hisi_acc_vfio_pci_migrn_state_ops = {
 	.migration_get_state = hisi_acc_vfio_pci_get_device_state,
 };
 
+static int hisi_acc_vfio_pci_migrn_init_dev(struct vfio_device *core_vdev)
+{
+	struct hisi_acc_vf_core_device *hisi_acc_vdev = container_of(core_vdev,
+			struct hisi_acc_vf_core_device, core_device.vdev);
+	struct pci_dev *pdev = to_pci_dev(core_vdev->dev);
+	struct hisi_qm *pf_qm = hisi_acc_get_pf_qm(pdev);
+
+	hisi_acc_vdev->vf_id = pci_iov_vf_id(pdev) + 1;
+	hisi_acc_vdev->pf_qm = pf_qm;
+	hisi_acc_vdev->vf_dev = pdev;
+	mutex_init(&hisi_acc_vdev->state_mutex);
+
+	core_vdev->migration_flags = VFIO_MIGRATION_STOP_COPY;
+	core_vdev->mig_ops = &hisi_acc_vfio_pci_migrn_state_ops;
+
+	return vfio_pci_core_init_dev(core_vdev);
+}
+
 static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = {
 	.name = "hisi-acc-vfio-pci-migration",
+	.init = hisi_acc_vfio_pci_migrn_init_dev,
+	.release = vfio_pci_core_release_dev,
 	.open_device = hisi_acc_vfio_pci_open_device,
 	.close_device = hisi_acc_vfio_pci_close_device,
 	.ioctl = hisi_acc_vfio_pci_ioctl,
@@ -1228,6 +1248,8 @@ static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = {
 
 static const struct vfio_device_ops hisi_acc_vfio_pci_ops = {
 	.name = "hisi-acc-vfio-pci",
+	.init = vfio_pci_core_init_dev,
+	.release = vfio_pci_core_release_dev,
 	.open_device = hisi_acc_vfio_pci_open_device,
 	.close_device = vfio_pci_core_close_device,
 	.ioctl = vfio_pci_core_ioctl,
@@ -1239,63 +1261,36 @@ static const struct vfio_device_ops hisi_acc_vfio_pci_ops = {
 	.match = vfio_pci_core_match,
 };
 
-static int
-hisi_acc_vfio_pci_migrn_init(struct hisi_acc_vf_core_device *hisi_acc_vdev,
-			     struct pci_dev *pdev, struct hisi_qm *pf_qm)
-{
-	int vf_id;
-
-	vf_id = pci_iov_vf_id(pdev);
-	if (vf_id < 0)
-		return vf_id;
-
-	hisi_acc_vdev->vf_id = vf_id + 1;
-	hisi_acc_vdev->core_device.vdev.migration_flags =
-					VFIO_MIGRATION_STOP_COPY;
-	hisi_acc_vdev->pf_qm = pf_qm;
-	hisi_acc_vdev->vf_dev = pdev;
-	mutex_init(&hisi_acc_vdev->state_mutex);
-
-	return 0;
-}
-
 static int hisi_acc_vfio_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	struct hisi_acc_vf_core_device *hisi_acc_vdev;
+	const struct vfio_device_ops *ops = &hisi_acc_vfio_pci_ops;
 	struct hisi_qm *pf_qm;
+	int vf_id;
 	int ret;
 
-	hisi_acc_vdev = kzalloc(sizeof(*hisi_acc_vdev), GFP_KERNEL);
-	if (!hisi_acc_vdev)
-		return -ENOMEM;
-
 	pf_qm = hisi_acc_get_pf_qm(pdev);
 	if (pf_qm && pf_qm->ver >= QM_HW_V3) {
-		ret = hisi_acc_vfio_pci_migrn_init(hisi_acc_vdev, pdev, pf_qm);
-		if (!ret) {
-			vfio_pci_core_init_device(&hisi_acc_vdev->core_device, pdev,
-						  &hisi_acc_vfio_pci_migrn_ops);
-			hisi_acc_vdev->core_device.vdev.mig_ops =
-					&hisi_acc_vfio_pci_migrn_state_ops;
-		} else {
+		vf_id = pci_iov_vf_id(pdev);
+		if (vf_id >= 0)
+			ops = &hisi_acc_vfio_pci_migrn_ops;
+		else
 			pci_warn(pdev, "migration support failed, continue with generic interface\n");
-			vfio_pci_core_init_device(&hisi_acc_vdev->core_device, pdev,
-						  &hisi_acc_vfio_pci_ops);
-		}
-	} else {
-		vfio_pci_core_init_device(&hisi_acc_vdev->core_device, pdev,
-					  &hisi_acc_vfio_pci_ops);
 	}
 
+	hisi_acc_vdev = vfio_alloc_device(hisi_acc_vf_core_device,
+					  core_device.vdev, &pdev->dev, ops);
+	if (IS_ERR(hisi_acc_vdev))
+		return PTR_ERR(hisi_acc_vdev);
+
 	dev_set_drvdata(&pdev->dev, &hisi_acc_vdev->core_device);
 	ret = vfio_pci_core_register_device(&hisi_acc_vdev->core_device);
 	if (ret)
-		goto out_free;
+		goto out_put_vdev;
 	return 0;
 
-out_free:
-	vfio_pci_core_uninit_device(&hisi_acc_vdev->core_device);
-	kfree(hisi_acc_vdev);
+out_put_vdev:
+	vfio_put_device(&hisi_acc_vdev->core_device.vdev);
 	return ret;
 }
 
@@ -1304,8 +1299,7 @@ static void hisi_acc_vfio_pci_remove(struct pci_dev *pdev)
 	struct hisi_acc_vf_core_device *hisi_acc_vdev = hisi_acc_drvdata(pdev);
 
 	vfio_pci_core_unregister_device(&hisi_acc_vdev->core_device);
-	vfio_pci_core_uninit_device(&hisi_acc_vdev->core_device);
-	kfree(hisi_acc_vdev);
+	vfio_put_device(&hisi_acc_vdev->core_device.vdev);
 }
 
 static const struct pci_device_id hisi_acc_vfio_pci_table[] = {
drivers/vfio/pci/vfio_pci_core.c +0 -30
index 77d33739c6e8..59a28251bb0b 100644
@@ -2113,36 +2113,6 @@ void vfio_pci_core_release_dev(struct vfio_device *core_vdev)
 }
 EXPORT_SYMBOL_GPL(vfio_pci_core_release_dev);
 
-void vfio_pci_core_init_device(struct vfio_pci_core_device *vdev,
-			       struct pci_dev *pdev,
-			       const struct vfio_device_ops *vfio_pci_ops)
-{
-	vfio_init_group_dev(&vdev->vdev, &pdev->dev, vfio_pci_ops);
-	vdev->pdev = pdev;
-	vdev->irq_type = VFIO_PCI_NUM_IRQS;
-	mutex_init(&vdev->igate);
-	spin_lock_init(&vdev->irqlock);
-	mutex_init(&vdev->ioeventfds_lock);
-	INIT_LIST_HEAD(&vdev->dummy_resources_list);
-	INIT_LIST_HEAD(&vdev->ioeventfds_list);
-	mutex_init(&vdev->vma_lock);
-	INIT_LIST_HEAD(&vdev->vma_list);
-	INIT_LIST_HEAD(&vdev->sriov_pfs_item);
-	init_rwsem(&vdev->memory_lock);
-}
-EXPORT_SYMBOL_GPL(vfio_pci_core_init_device);
-
-void vfio_pci_core_uninit_device(struct vfio_pci_core_device *vdev)
-{
-	mutex_destroy(&vdev->igate);
-	mutex_destroy(&vdev->ioeventfds_lock);
-	mutex_destroy(&vdev->vma_lock);
-	vfio_uninit_group_dev(&vdev->vdev);
-	kfree(vdev->region);
-	kfree(vdev->pm_save);
-}
-EXPORT_SYMBOL_GPL(vfio_pci_core_uninit_device);
-
 int vfio_pci_core_register_device(struct vfio_pci_core_device *vdev)
 {
 	struct pci_dev *pdev = vdev->pdev;
include/linux/vfio_pci_core.h +0 -4
index 0499ea836058..367fd79226a3 100644
@@ -106,13 +106,9 @@ int vfio_pci_core_register_dev_region(struct vfio_pci_core_device *vdev,
 void vfio_pci_core_set_params(bool nointxmask, bool is_disable_vga,
 			      bool is_disable_idle_d3);
 void vfio_pci_core_close_device(struct vfio_device *core_vdev);
-void vfio_pci_core_init_device(struct vfio_pci_core_device *vdev,
-			       struct pci_dev *pdev,
-			       const struct vfio_device_ops *vfio_pci_ops);
 int vfio_pci_core_init_dev(struct vfio_device *core_vdev);
 void vfio_pci_core_release_dev(struct vfio_device *core_vdev);
 int vfio_pci_core_register_device(struct vfio_pci_core_device *vdev);
-void vfio_pci_core_uninit_device(struct vfio_pci_core_device *vdev);
 void vfio_pci_core_unregister_device(struct vfio_pci_core_device *vdev);
 extern const struct pci_error_handlers vfio_pci_core_err_handlers;
 int vfio_pci_core_sriov_configure(struct vfio_pci_core_device *vdev,
Signature · 1 line
2.21.3
author reviewer
#8 lore raw

PATCH v3 5/15 vfio/mdpy: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

and manage mdpy_count inside @init/@release.

---
 samples/vfio-mdev/mdpy.c | 81 +++++++++++++++++++++++-----------------
 1 file changed, 47 insertions(+), 34 deletions(-)
samples/vfio-mdev/mdpy.c +47 -34
index e8c46eb2e246..a07dac16d873 100644
@@ -216,61 +216,77 @@ static int mdpy_reset(struct mdev_state *mdev_state)
 	return 0;
 }
 
-static int mdpy_probe(struct mdev_device *mdev)
+static int mdpy_init_dev(struct vfio_device *vdev)
 {
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+	struct mdev_device *mdev = to_mdev_device(vdev->dev);
 	const struct mdpy_type *type =
 		&mdpy_types[mdev_get_type_group_id(mdev)];
-	struct device *dev = mdev_dev(mdev);
-	struct mdev_state *mdev_state;
 	u32 fbsize;
-	int ret;
+	int ret = -ENOMEM;
 
 	if (mdpy_count >= max_devices)
-		return -ENOMEM;
-
-	mdev_state = kzalloc(sizeof(struct mdev_state), GFP_KERNEL);
-	if (mdev_state == NULL)
-		return -ENOMEM;
-	vfio_init_group_dev(&mdev_state->vdev, &mdev->dev, &mdpy_dev_ops);
+		return ret;
 
 	mdev_state->vconfig = kzalloc(MDPY_CONFIG_SPACE_SIZE, GFP_KERNEL);
-	if (mdev_state->vconfig == NULL) {
-		ret = -ENOMEM;
-		goto err_state;
-	}
+	if (!mdev_state->vconfig)
+		return ret;
 
 	fbsize = roundup_pow_of_two(type->width * type->height * type->bytepp);
 
 	mdev_state->memblk = vmalloc_user(fbsize);
-	if (!mdev_state->memblk) {
-		ret = -ENOMEM;
-		goto err_vconfig;
-	}
-	dev_info(dev, "%s: %s (%dx%d)\n", __func__, type->name, type->width,
-		 type->height);
+	if (!mdev_state->memblk)
+		goto out_vconfig;
 
 	mutex_init(&mdev_state->ops_lock);
 	mdev_state->mdev = mdev;
-	mdev_state->type    = type;
+	mdev_state->type = type;
 	mdev_state->memsize = fbsize;
 	mdpy_create_config_space(mdev_state);
 	mdpy_reset(mdev_state);
 
+	dev_info(vdev->dev, "%s: %s (%dx%d)\n", __func__, type->name, type->width,
+		 type->height);
+
 	mdpy_count++;
+	return 0;
+
+out_vconfig:
+	kfree(mdev_state->vconfig);
+	return ret;
+}
+
+static int mdpy_probe(struct mdev_device *mdev)
+{
+	struct mdev_state *mdev_state;
+	int ret;
+
+	mdev_state = vfio_alloc_device(mdev_state, vdev, &mdev->dev,
+				       &mdpy_dev_ops);
+	if (IS_ERR(mdev_state))
+		return PTR_ERR(mdev_state);
 
 	ret = vfio_register_emulated_iommu_dev(&mdev_state->vdev);
 	if (ret)
-		goto err_mem;
+		goto err_put_vdev;
 	dev_set_drvdata(&mdev->dev, mdev_state);
 	return 0;
-err_mem:
+
+err_put_vdev:
+	vfio_put_device(&mdev_state->vdev);
+	return ret;
+}
+
+static void mdpy_release_dev(struct vfio_device *vdev)
+{
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+
 	vfree(mdev_state->memblk);
-err_vconfig:
 	kfree(mdev_state->vconfig);
-err_state:
-	vfio_uninit_group_dev(&mdev_state->vdev);
-	kfree(mdev_state);
-	return ret;
+	vfio_free_device(vdev);
+	mdpy_count--;
 }
 
 static void mdpy_remove(struct mdev_device *mdev)
@@ -280,12 +296,7 @@ static void mdpy_remove(struct mdev_device *mdev)
 	dev_info(&mdev->dev, "%s\n", __func__);
 
 	vfio_unregister_group_dev(&mdev_state->vdev);
-	vfree(mdev_state->memblk);
-	kfree(mdev_state->vconfig);
-	vfio_uninit_group_dev(&mdev_state->vdev);
-	kfree(mdev_state);
-
-	mdpy_count--;
+	vfio_put_device(&mdev_state->vdev);
 }
 
 static ssize_t mdpy_read(struct vfio_device *vdev, char __user *buf,
@@ -708,6 +719,8 @@ static struct attribute_group *mdev_type_groups[] = {
 };
 
 static const struct vfio_device_ops mdpy_dev_ops = {
+	.init = mdpy_init_dev,
+	.release = mdpy_release_dev,
 	.read = mdpy_read,
 	.write = mdpy_write,
 	.ioctl = mdpy_ioctl,
Signature · 1 line
2.21.3
author reviewer
#9 lore raw

PATCH v3 6/15 vfio/mtty: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

and manage available ports inside @init/@release.

---
 samples/vfio-mdev/mtty.c | 67 +++++++++++++++++++++++-----------------
 1 file changed, 39 insertions(+), 28 deletions(-)
samples/vfio-mdev/mtty.c +39 -28
index f42a59ed2e3f..41301d50b247 100644
@@ -703,9 +703,11 @@ static ssize_t mdev_access(struct mdev_state *mdev_state, u8 *buf, size_t count,
 	return ret;
 }
 
-static int mtty_probe(struct mdev_device *mdev)
+static int mtty_init_dev(struct vfio_device *vdev)
 {
-	struct mdev_state *mdev_state;
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+	struct mdev_device *mdev = to_mdev_device(vdev->dev);
 	int nr_ports = mdev_get_type_group_id(mdev) + 1;
 	int avail_ports = atomic_read(&mdev_avail_ports);
 	int ret;
@@ -716,58 +718,65 @@ static int mtty_probe(struct mdev_device *mdev)
 	} while (!atomic_try_cmpxchg(&mdev_avail_ports,
 				     &avail_ports, avail_ports - nr_ports));
 
-	mdev_state = kzalloc(sizeof(struct mdev_state), GFP_KERNEL);
-	if (mdev_state == NULL) {
-		ret = -ENOMEM;
-		goto err_nr_ports;
-	}
-
-	vfio_init_group_dev(&mdev_state->vdev, &mdev->dev, &mtty_dev_ops);
-
 	mdev_state->nr_ports = nr_ports;
 	mdev_state->irq_index = -1;
 	mdev_state->s[0].max_fifo_size = MAX_FIFO_SIZE;
 	mdev_state->s[1].max_fifo_size = MAX_FIFO_SIZE;
 	mutex_init(&mdev_state->rxtx_lock);
-	mdev_state->vconfig = kzalloc(MTTY_CONFIG_SPACE_SIZE, GFP_KERNEL);
 
-	if (mdev_state->vconfig == NULL) {
+	mdev_state->vconfig = kzalloc(MTTY_CONFIG_SPACE_SIZE, GFP_KERNEL);
+	if (!mdev_state->vconfig) {
 		ret = -ENOMEM;
-		goto err_state;
+		goto err_nr_ports;
 	}
 
 	mutex_init(&mdev_state->ops_lock);
 	mdev_state->mdev = mdev;
-
 	mtty_create_config_space(mdev_state);
+	return 0;
+
+err_nr_ports:
+	atomic_add(nr_ports, &mdev_avail_ports);
+	return ret;
+}
+
+static int mtty_probe(struct mdev_device *mdev)
+{
+	struct mdev_state *mdev_state;
+	int ret;
+
+	mdev_state = vfio_alloc_device(mdev_state, vdev, &mdev->dev,
+				       &mtty_dev_ops);
+	if (IS_ERR(mdev_state))
+		return PTR_ERR(mdev_state);
 
 	ret = vfio_register_emulated_iommu_dev(&mdev_state->vdev);
 	if (ret)
-		goto err_vconfig;
+		goto err_put_vdev;
 	dev_set_drvdata(&mdev->dev, mdev_state);
 	return 0;
 
-err_vconfig:
-	kfree(mdev_state->vconfig);
-err_state:
-	vfio_uninit_group_dev(&mdev_state->vdev);
-	kfree(mdev_state);
-err_nr_ports:
-	atomic_add(nr_ports, &mdev_avail_ports);
+err_put_vdev:
+	vfio_put_device(&mdev_state->vdev);
 	return ret;
 }
 
+static void mtty_release_dev(struct vfio_device *vdev)
+{
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+
+	kfree(mdev_state->vconfig);
+	vfio_free_device(vdev);
+	atomic_add(mdev_state->nr_ports, &mdev_avail_ports);
+}
+
 static void mtty_remove(struct mdev_device *mdev)
 {
 	struct mdev_state *mdev_state = dev_get_drvdata(&mdev->dev);
-	int nr_ports = mdev_state->nr_ports;
 
 	vfio_unregister_group_dev(&mdev_state->vdev);
-
-	kfree(mdev_state->vconfig);
-	vfio_uninit_group_dev(&mdev_state->vdev);
-	kfree(mdev_state);
-	atomic_add(nr_ports, &mdev_avail_ports);
+	vfio_put_device(&mdev_state->vdev);
 }
 
 static int mtty_reset(struct mdev_state *mdev_state)
@@ -1287,6 +1296,8 @@ static struct attribute_group *mdev_type_groups[] = {
 
 static const struct vfio_device_ops mtty_dev_ops = {
 	.name = "vfio-mtty",
+	.init = mtty_init_dev,
+	.release = mtty_release_dev,
 	.read = mtty_read,
 	.write = mtty_write,
 	.ioctl = mtty_ioctl,
Signature · 1 line
2.21.3
#10 lore raw

Re: [PATCH v3 06/15] vfio/mtty: Use the new device life cycle helpers

Quoted 15 lines · On Fri, 9 Sep 2022 18:22:38 +0800 Kevin Tian <[email protected]> wrote:

From: Yi Liu <[email protected]>

and manage available ports inside @init/@release.

---
 samples/vfio-mdev/mtty.c | 67 +++++++++++++++++++++++-----------------
 1 file changed, 39 insertions(+), 28 deletions(-)
samples/vfio-mdev/mtty.c +0 -0
index f42a59ed2e3f..41301d50b247 100644

...

Quoted 27 lines
samples/vfio-mdev/mtty.c +12 -8
+static int mtty_probe(struct mdev_device *mdev)
+{
+	struct mdev_state *mdev_state;
+	int ret;
+
+	mdev_state = vfio_alloc_device(mdev_state, vdev, &mdev->dev,
+				       &mtty_dev_ops);
+	if (IS_ERR(mdev_state))
+		return PTR_ERR(mdev_state);
 
 	ret = vfio_register_emulated_iommu_dev(&mdev_state->vdev);
 	if (ret)
-		goto err_vconfig;
+		goto err_put_vdev;
 	dev_set_drvdata(&mdev->dev, mdev_state);
 	return 0;
 
-err_vconfig:
-	kfree(mdev_state->vconfig);
-err_state:
-	vfio_uninit_group_dev(&mdev_state->vdev);
-	kfree(mdev_state);
-err_nr_ports:
-	atomic_add(nr_ports, &mdev_avail_ports);
+err_put_vdev:
+	vfio_put_device(&mdev_state->vdev);
 	return ret;

samples/vfio-mdev/mtty.c static int mtty_probe(struct mdev_device *mdev) · line 771

Quoted 10 lines
samples/vfio-mdev/mtty.c +8 -0
 }
 
+static void mtty_release_dev(struct vfio_device *vdev)
+{
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+
+	kfree(mdev_state->vconfig);
+	vfio_free_device(vdev);
+	atomic_add(mdev_state->nr_ports, &mdev_avail_ports);

I must be missing something, isn't this a use-after-free?

mdev_state is allocated via vfio_alloc_device(), where vdev is the first entry in that structure, so this is equivalent to kvfree(mdev_state). mbochs has the same issue. mdpy and vfio-ap adjust global counters after vfio_free_device(), which I think muddies the situation. Shouldn't we look suspiciously at any .release callback where vfio_free_device() isn't the last thing executed? Thanks,

Alex

author reviewer
#11 lore raw

RE: [PATCH v3 06/15] vfio/mtty: Use the new device life cycle helpers

Quoted 51 lines

From: Alex Williamson <[email protected]>
Sent: Wednesday, September 21, 2022 3:17 AM

Quoted 15 lines · On Fri, 9 Sep 2022 18:22:38 +0800 Kevin Tian <[email protected]> wrote:

From: Yi Liu <[email protected]>

and manage available ports inside @init/@release.

---
 samples/vfio-mdev/mtty.c | 67 +++++++++++++++++++++++-----------------
 1 file changed, 39 insertions(+), 28 deletions(-)
samples/vfio-mdev/mtty.c +0 -0
index f42a59ed2e3f..41301d50b247 100644

...

Quoted 29 lines
samples/vfio-mdev/mtty.c +12 -8
+static int mtty_probe(struct mdev_device *mdev)
+{
+	struct mdev_state *mdev_state;
+	int ret;
+
+	mdev_state = vfio_alloc_device(mdev_state, vdev, &mdev->dev,
+				       &mtty_dev_ops);
+	if (IS_ERR(mdev_state))
+		return PTR_ERR(mdev_state);
 
 	ret = vfio_register_emulated_iommu_dev(&mdev_state->vdev);
 	if (ret)
-		goto err_vconfig;
+		goto err_put_vdev;
 	dev_set_drvdata(&mdev->dev, mdev_state);
 	return 0;
 
-err_vconfig:
-	kfree(mdev_state->vconfig);
-err_state:
-	vfio_uninit_group_dev(&mdev_state->vdev);
-	kfree(mdev_state);
-err_nr_ports:
-	atomic_add(nr_ports, &mdev_avail_ports);
+err_put_vdev:
+	vfio_put_device(&mdev_state->vdev);
 	return ret;
 }
Quoted 10 lines

samples/vfio-mdev/mtty.c static int mtty_probe(struct mdev_device *mdev) · line 771

Quoted 8 lines
samples/vfio-mdev/mtty.c +8 -0
+static void mtty_release_dev(struct vfio_device *vdev)
+{
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+
+	kfree(mdev_state->vconfig);
+	vfio_free_device(vdev);
+	atomic_add(mdev_state->nr_ports, &mdev_avail_ports);

I must be missing something, isn't this a use-after-free?

Yes, it's a use-after-free indeed. Thanks for catching it!

Quoted 8 lines

mdev_state is allocated via vfio_alloc_device(), where vdev is the first entry in that structure, so this is equivalent to kvfree(mdev_state). mbochs has the same issue. mdpy and vfio-ap adjust global counters after vfio_free_device(), which I think muddies the situation. Shouldn't we look suspiciously at any .release callback where vfio_free_device() isn't the last thing executed? Thanks,

Yes. I'll scrutinize it again. To be consistent I'll make sure the free is the last line in all .release callbacks though not required for some (e.g. mdpy and vfio-ap).

author reviewer
#12 lore raw

PATCH v3 7/15 vfio/mbochs: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

and manage avail_mbytes inside @init/@release.

---
 samples/vfio-mdev/mbochs.c | 73 ++++++++++++++++++++++++--------------
 1 file changed, 46 insertions(+), 27 deletions(-)
samples/vfio-mdev/mbochs.c +46 -27
index 344c2901a82b..df95f25fbc0e 100644
@@ -505,13 +505,14 @@ static int mbochs_reset(struct mdev_state *mdev_state)
 	return 0;
 }
 
-static int mbochs_probe(struct mdev_device *mdev)
+static int mbochs_init_dev(struct vfio_device *vdev)
 {
-	int avail_mbytes = atomic_read(&mbochs_avail_mbytes);
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+	struct mdev_device *mdev = to_mdev_device(vdev->dev);
 	const struct mbochs_type *type =
 		&mbochs_types[mdev_get_type_group_id(mdev)];
-	struct device *dev = mdev_dev(mdev);
-	struct mdev_state *mdev_state;
+	int avail_mbytes = atomic_read(&mbochs_avail_mbytes);
 	int ret = -ENOMEM;
 
 	do {
@@ -520,14 +521,9 @@ static int mbochs_probe(struct mdev_device *mdev)
 	} while (!atomic_try_cmpxchg(&mbochs_avail_mbytes, &avail_mbytes,
 				     avail_mbytes - type->mbytes));
 
-	mdev_state = kzalloc(sizeof(struct mdev_state), GFP_KERNEL);
-	if (mdev_state == NULL)
-		goto err_avail;
-	vfio_init_group_dev(&mdev_state->vdev, &mdev->dev, &mbochs_dev_ops);
-
 	mdev_state->vconfig = kzalloc(MBOCHS_CONFIG_SPACE_SIZE, GFP_KERNEL);
-	if (mdev_state->vconfig == NULL)
-		goto err_mem;
+	if (!mdev_state->vconfig)
+		goto err_avail;
 
 	mdev_state->memsize = type->mbytes * 1024 * 1024;
 	mdev_state->pagecount = mdev_state->memsize >> PAGE_SHIFT;
@@ -535,10 +531,7 @@ static int mbochs_probe(struct mdev_device *mdev)
 				    sizeof(struct page *),
 				    GFP_KERNEL);
 	if (!mdev_state->pages)
-		goto err_mem;
-
-	dev_info(dev, "%s: %s, %d MB, %ld pages\n", __func__,
-		 type->name, type->mbytes, mdev_state->pagecount);
+		goto err_vconfig;
 
 	mutex_init(&mdev_state->ops_lock);
 	mdev_state->mdev = mdev;
@@ -553,19 +546,47 @@ static int mbochs_probe(struct mdev_device *mdev)
 	mbochs_create_config_space(mdev_state);
 	mbochs_reset(mdev_state);
 
+	dev_info(vdev->dev, "%s: %s, %d MB, %ld pages\n", __func__,
+		 type->name, type->mbytes, mdev_state->pagecount);
+	return 0;
+
+err_vconfig:
+	kfree(mdev_state->vconfig);
+err_avail:
+	atomic_add(type->mbytes, &mbochs_avail_mbytes);
+	return ret;
+}
+
+static int mbochs_probe(struct mdev_device *mdev)
+{
+	struct mdev_state *mdev_state;
+	int ret = -ENOMEM;
+
+	mdev_state = vfio_alloc_device(mdev_state, vdev, &mdev->dev,
+				       &mbochs_dev_ops);
+	if (IS_ERR(mdev_state))
+		return PTR_ERR(mdev_state);
+
 	ret = vfio_register_emulated_iommu_dev(&mdev_state->vdev);
 	if (ret)
-		goto err_mem;
+		goto err_put_vdev;
 	dev_set_drvdata(&mdev->dev, mdev_state);
 	return 0;
-err_mem:
-	vfio_uninit_group_dev(&mdev_state->vdev);
+
+err_put_vdev:
+	vfio_put_device(&mdev_state->vdev);
+	return ret;
+}
+
+static void mbochs_release_dev(struct vfio_device *vdev)
+{
+	struct mdev_state *mdev_state =
+		container_of(vdev, struct mdev_state, vdev);
+
 	kfree(mdev_state->pages);
 	kfree(mdev_state->vconfig);
-	kfree(mdev_state);
-err_avail:
-	atomic_add(type->mbytes, &mbochs_avail_mbytes);
-	return ret;
+	vfio_free_device(vdev);
+	atomic_add(mdev_state->type->mbytes, &mbochs_avail_mbytes);
 }
 
 static void mbochs_remove(struct mdev_device *mdev)
@@ -573,11 +594,7 @@ static void mbochs_remove(struct mdev_device *mdev)
 	struct mdev_state *mdev_state = dev_get_drvdata(&mdev->dev);
 
 	vfio_unregister_group_dev(&mdev_state->vdev);
-	vfio_uninit_group_dev(&mdev_state->vdev);
-	atomic_add(mdev_state->type->mbytes, &mbochs_avail_mbytes);
-	kfree(mdev_state->pages);
-	kfree(mdev_state->vconfig);
-	kfree(mdev_state);
+	vfio_put_device(&mdev_state->vdev);
 }
 
 static ssize_t mbochs_read(struct vfio_device *vdev, char __user *buf,
@@ -1397,6 +1414,8 @@ static struct attribute_group *mdev_type_groups[] = {
 
 static const struct vfio_device_ops mbochs_dev_ops = {
 	.close_device = mbochs_close_device,
+	.init = mbochs_init_dev,
+	.release = mbochs_release_dev,
 	.read = mbochs_read,
 	.write = mbochs_write,
 	.ioctl = mbochs_ioctl,
Signature · 1 line
2.21.3
author reviewer
#13 lore raw

PATCH v3 8/15 drm/i915/gvt: Use the new device life cycle helpers

In reply to Kevin Tian

Move vfio_device to the start of intel_vgpu as required by the new helpers.

Change intel_gvt_create_vgpu() to use intel_vgpu as the first param as other vgpu helpers do.

---
 drivers/gpu/drm/i915/gvt/gvt.h   |  5 ++-
 drivers/gpu/drm/i915/gvt/kvmgt.c | 52 ++++++++++++++++++++++----------
 drivers/gpu/drm/i915/gvt/vgpu.c  | 33 ++++++++------------
 3 files changed, 50 insertions(+), 40 deletions(-)
drivers/gpu/drm/i915/gvt/gvt.h +2 -3
index 705689e64011..89fab7896fc6 100644
@@ -172,6 +172,7 @@ struct intel_vgpu_submission {
 #define KVMGT_DEBUGFS_FILENAME		"kvmgt_nr_cache_entries"
 
 struct intel_vgpu {
+	struct vfio_device vfio_device;
 	struct intel_gvt *gvt;
 	struct mutex vgpu_lock;
 	int id;
@@ -211,7 +212,6 @@ struct intel_vgpu {
 
 	u32 scan_nonprivbb;
 
-	struct vfio_device vfio_device;
 	struct vfio_region *region;
 	int num_regions;
 	struct eventfd_ctx *intx_trigger;
@@ -494,8 +494,7 @@ void intel_gvt_clean_vgpu_types(struct intel_gvt *gvt);
 
 struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt);
 void intel_gvt_destroy_idle_vgpu(struct intel_vgpu *vgpu);
-struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt,
-					 struct intel_vgpu_type *type);
+int intel_gvt_create_vgpu(struct intel_vgpu *vgpu, struct intel_vgpu_type *type);
 void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu);
 void intel_gvt_release_vgpu(struct intel_vgpu *vgpu);
 void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
drivers/gpu/drm/i915/gvt/kvmgt.c +36 -16

85 lines not on this page · raw message

drivers/gpu/drm/i915/gvt/vgpu.c +12 -21
index 46da19b3225d..5c533fbc2c8d 100644
@@ -302,8 +302,6 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
 	mutex_lock(&gvt->lock);
 	intel_gvt_update_vgpu_types(gvt);
 	mutex_unlock(&gvt->lock);
-
-	vfree(vgpu);
 }
 
 #define IDLE_VGPU_IDR 0
@@ -363,28 +361,23 @@ void intel_gvt_destroy_idle_vgpu(struct intel_vgpu *vgpu)
 	vfree(vgpu);
 }
 
-static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
-		struct intel_vgpu_creation_params *param)
+static int __intel_gvt_create_vgpu(struct intel_vgpu *vgpu,
+				   struct intel_vgpu_creation_params *param)
 {
+	struct intel_gvt *gvt = vgpu->gvt;
 	struct drm_i915_private *dev_priv = gvt->gt->i915;
-	struct intel_vgpu *vgpu;
 	int ret;
 
 	gvt_dbg_core("low %llu MB high %llu MB fence %llu\n",
 			param->low_gm_sz, param->high_gm_sz,
 			param->fence_sz);
 
-	vgpu = vzalloc(sizeof(*vgpu));
-	if (!vgpu)
-		return ERR_PTR(-ENOMEM);
-
 	ret = idr_alloc(&gvt->vgpu_idr, vgpu, IDLE_VGPU_IDR + 1, GVT_MAX_VGPU,
 		GFP_KERNEL);
 	if (ret < 0)
-		goto out_free_vgpu;
+		return ret;
 
 	vgpu->id = ret;
-	vgpu->gvt = gvt;
 	vgpu->sched_ctl.weight = param->weight;
 	mutex_init(&vgpu->vgpu_lock);
 	mutex_init(&vgpu->dmabuf_lock);
@@ -437,7 +430,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	if (ret)
 		goto out_clean_sched_policy;
 
-	return vgpu;
+	return 0;
 
 out_clean_sched_policy:
 	intel_vgpu_clean_sched_policy(vgpu);
@@ -455,9 +448,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	intel_vgpu_clean_mmio(vgpu);
 out_clean_idr:
 	idr_remove(&gvt->vgpu_idr, vgpu->id);
-out_free_vgpu:
-	vfree(vgpu);
-	return ERR_PTR(ret);
+	return ret;
 }
 
 /**
@@ -470,11 +461,11 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
  * Returns:
  * pointer to intel_vgpu, error pointer if failed.
  */
-struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt,
-				struct intel_vgpu_type *type)
+int intel_gvt_create_vgpu(struct intel_vgpu *vgpu, struct intel_vgpu_type *type)
 {
+	struct intel_gvt *gvt = vgpu->gvt;
 	struct intel_vgpu_creation_params param;
-	struct intel_vgpu *vgpu;
+	int ret;
 
 	param.primary = 1;
 	param.low_gm_sz = type->low_gm_size;
@@ -488,15 +479,15 @@ struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	param.high_gm_sz = BYTES_TO_MB(param.high_gm_sz);
 
 	mutex_lock(&gvt->lock);
-	vgpu = __intel_gvt_create_vgpu(gvt, &param);
-	if (!IS_ERR(vgpu)) {
+	ret = __intel_gvt_create_vgpu(vgpu, &param);
+	if (!ret) {
 		/* calculate left instance change for types */
 		intel_gvt_update_vgpu_types(gvt);
 		intel_gvt_update_reg_whitelist(vgpu);
 	}
 	mutex_unlock(&gvt->lock);
 
-	return vgpu;
+	return ret;
 }
 
 /**
Signature · 1 line
2.21.3
author reviewer
#14 lore raw

PATCH v3 9/15 vfio/ap: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

and manage available_instances inside @init/@release.

---
 drivers/s390/crypto/vfio_ap_ops.c | 50 ++++++++++++++++++-------------
 1 file changed, 29 insertions(+), 21 deletions(-)
drivers/s390/crypto/vfio_ap_ops.c +29 -21
index 6c8c41fac4e1..803aadfd0876 100644
@@ -684,42 +684,44 @@ static bool vfio_ap_mdev_filter_matrix(unsigned long *apm, unsigned long *aqm,
 			     AP_DOMAINS);
 }
 
-static int vfio_ap_mdev_probe(struct mdev_device *mdev)
+static int vfio_ap_mdev_init_dev(struct vfio_device *vdev)
 {
-	struct ap_matrix_mdev *matrix_mdev;
-	int ret;
+	struct ap_matrix_mdev *matrix_mdev =
+		container_of(vdev, struct ap_matrix_mdev, vdev);
 
 	if ((atomic_dec_if_positive(&matrix_dev->available_instances) < 0))
 		return -EPERM;
 
-	matrix_mdev = kzalloc(sizeof(*matrix_mdev), GFP_KERNEL);
-	if (!matrix_mdev) {
-		ret = -ENOMEM;
-		goto err_dec_available;
-	}
-	vfio_init_group_dev(&matrix_mdev->vdev, &mdev->dev,
-			    &vfio_ap_matrix_dev_ops);
-
-	matrix_mdev->mdev = mdev;
+	matrix_mdev->mdev = to_mdev_device(vdev->dev);
 	vfio_ap_matrix_init(&matrix_dev->info, &matrix_mdev->matrix);
 	matrix_mdev->pqap_hook = handle_pqap;
 	vfio_ap_matrix_init(&matrix_dev->info, &matrix_mdev->shadow_apcb);
 	hash_init(matrix_mdev->qtable.queues);
 
+	return 0;
+}
+
+static int vfio_ap_mdev_probe(struct mdev_device *mdev)
+{
+	struct ap_matrix_mdev *matrix_mdev;
+	int ret;
+
+	matrix_mdev = vfio_alloc_device(ap_matrix_mdev, vdev, &mdev->dev,
+					&vfio_ap_matrix_dev_ops);
+	if (IS_ERR(matrix_mdev))
+		return PTR_ERR(matrix_mdev);
+
 	ret = vfio_register_emulated_iommu_dev(&matrix_mdev->vdev);
 	if (ret)
-		goto err_list;
+		goto err_put_vdev;
 	dev_set_drvdata(&mdev->dev, matrix_mdev);
 	mutex_lock(&matrix_dev->mdevs_lock);
 	list_add(&matrix_mdev->node, &matrix_dev->mdev_list);
 	mutex_unlock(&matrix_dev->mdevs_lock);
 	return 0;
 
-err_list:
-	vfio_uninit_group_dev(&matrix_mdev->vdev);
-	kfree(matrix_mdev);
-err_dec_available:
-	atomic_inc(&matrix_dev->available_instances);
+err_put_vdev:
+	vfio_put_device(&matrix_mdev->vdev);
 	return ret;
 }
 
@@ -766,6 +768,12 @@ static void vfio_ap_mdev_unlink_fr_queues(struct ap_matrix_mdev *matrix_mdev)
 	}
 }
 
+static void vfio_ap_mdev_release_dev(struct vfio_device *vdev)
+{
+	vfio_free_device(vdev);
+	atomic_inc(&matrix_dev->available_instances);
+}
+
 static void vfio_ap_mdev_remove(struct mdev_device *mdev)
 {
 	struct ap_matrix_mdev *matrix_mdev = dev_get_drvdata(&mdev->dev);
@@ -779,9 +787,7 @@ static void vfio_ap_mdev_remove(struct mdev_device *mdev)
 	list_del(&matrix_mdev->node);
 	mutex_unlock(&matrix_dev->mdevs_lock);
 	mutex_unlock(&matrix_dev->guests_lock);
-	vfio_uninit_group_dev(&matrix_mdev->vdev);
-	kfree(matrix_mdev);
-	atomic_inc(&matrix_dev->available_instances);
+	vfio_put_device(&matrix_mdev->vdev);
 }
 
 static ssize_t name_show(struct mdev_type *mtype,
@@ -1794,6 +1800,8 @@ static const struct attribute_group vfio_queue_attr_group = {
 };
 
 static const struct vfio_device_ops vfio_ap_matrix_dev_ops = {
+	.init = vfio_ap_mdev_init_dev,
+	.release = vfio_ap_mdev_release_dev,
 	.open_device = vfio_ap_mdev_open_device,
 	.close_device = vfio_ap_mdev_close_device,
 	.ioctl = vfio_ap_mdev_ioctl,
Signature · 1 line
2.21.3
author reviewer
#15 lore raw

PATCH v3 10/15 vfio/fsl-mc: Use the new device life cycle helpers

In reply to Kevin Tian

From: Yi Liu <[email protected]>

Also add a comment to mark that vfio core releases device_set if @init fails.

---
 drivers/vfio/fsl-mc/vfio_fsl_mc.c | 85 ++++++++++++++++++-------------
 1 file changed, 49 insertions(+), 36 deletions(-)
drivers/vfio/fsl-mc/vfio_fsl_mc.c +49 -36
index 42b344bd7cd5..b16874e913e4 100644
@@ -418,16 +418,7 @@ static int vfio_fsl_mc_mmap(struct vfio_device *core_vdev,
 	return vfio_fsl_mc_mmap_mmio(vdev->regions[index], vma);
 }
 
-static const struct vfio_device_ops vfio_fsl_mc_ops = {
-	.name		= "vfio-fsl-mc",
-	.open_device	= vfio_fsl_mc_open_device,
-	.close_device	= vfio_fsl_mc_close_device,
-	.ioctl		= vfio_fsl_mc_ioctl,
-	.read		= vfio_fsl_mc_read,
-	.write		= vfio_fsl_mc_write,
-	.mmap		= vfio_fsl_mc_mmap,
-};
-
+static const struct vfio_device_ops vfio_fsl_mc_ops;
 static int vfio_fsl_mc_bus_notifier(struct notifier_block *nb,
 				    unsigned long action, void *data)
 {
@@ -518,35 +509,43 @@ static void vfio_fsl_uninit_device(struct vfio_fsl_mc_device *vdev)
 	bus_unregister_notifier(&fsl_mc_bus_type, &vdev->nb);
 }
 
-static int vfio_fsl_mc_probe(struct fsl_mc_device *mc_dev)
+static int vfio_fsl_mc_init_dev(struct vfio_device *core_vdev)
 {
-	struct vfio_fsl_mc_device *vdev;
-	struct device *dev = &mc_dev->dev;
+	struct vfio_fsl_mc_device *vdev =
+		container_of(core_vdev, struct vfio_fsl_mc_device, vdev);
+	struct fsl_mc_device *mc_dev = to_fsl_mc_device(core_vdev->dev);
 	int ret;
 
-	vdev = kzalloc(sizeof(*vdev), GFP_KERNEL);
-	if (!vdev)
-		return -ENOMEM;
-
-	vfio_init_group_dev(&vdev->vdev, dev, &vfio_fsl_mc_ops);
 	vdev->mc_dev = mc_dev;
 	mutex_init(&vdev->igate);
 
 	if (is_fsl_mc_bus_dprc(mc_dev))
-		ret = vfio_assign_device_set(&vdev->vdev, &mc_dev->dev);
+		ret = vfio_assign_device_set(core_vdev, &mc_dev->dev);
 	else
-		ret = vfio_assign_device_set(&vdev->vdev, mc_dev->dev.parent);
-	if (ret)
-		goto out_uninit;
+		ret = vfio_assign_device_set(core_vdev, mc_dev->dev.parent);
 
-	ret = vfio_fsl_mc_init_device(vdev);
 	if (ret)
-		goto out_uninit;
+		return ret;
+
+	/* device_set is released by vfio core if @init fails */
+	return vfio_fsl_mc_init_device(vdev);
+}
+
+static int vfio_fsl_mc_probe(struct fsl_mc_device *mc_dev)
+{
+	struct vfio_fsl_mc_device *vdev;
+	struct device *dev = &mc_dev->dev;
+	int ret;
+
+	vdev = vfio_alloc_device(vfio_fsl_mc_device, vdev, dev,
+				 &vfio_fsl_mc_ops);
+	if (IS_ERR(vdev))
+		return PTR_ERR(vdev);
 
 	ret = vfio_register_group_dev(&vdev->vdev);
 	if (ret) {
 		dev_err(dev, "VFIO_FSL_MC: Failed to add to vfio group\n");
-		goto out_device;
+		goto out_put_vdev;
 	}
 
 	ret = vfio_fsl_mc_scan_container(mc_dev);
@@ -557,30 +556,44 @@ static int vfio_fsl_mc_probe(struct fsl_mc_device *mc_dev)
 
 out_group_dev:
 	vfio_unregister_group_dev(&vdev->vdev);
-out_device:
-	vfio_fsl_uninit_device(vdev);
-out_uninit:
-	vfio_uninit_group_dev(&vdev->vdev);
-	kfree(vdev);
+out_put_vdev:
+	vfio_put_device(&vdev->vdev);
 	return ret;
 }
 
+static void vfio_fsl_mc_release_dev(struct vfio_device *core_vdev)
+{
+	struct vfio_fsl_mc_device *vdev =
+		container_of(core_vdev, struct vfio_fsl_mc_device, vdev);
+
+	vfio_fsl_uninit_device(vdev);
+	mutex_destroy(&vdev->igate);
+	vfio_free_device(core_vdev);
+}
+
 static int vfio_fsl_mc_remove(struct fsl_mc_device *mc_dev)
 {
 	struct device *dev = &mc_dev->dev;
 	struct vfio_fsl_mc_device *vdev = dev_get_drvdata(dev);
 
 	vfio_unregister_group_dev(&vdev->vdev);
-	mutex_destroy(&vdev->igate);
-
 	dprc_remove_devices(mc_dev, NULL, 0);
-	vfio_fsl_uninit_device(vdev);
-
-	vfio_uninit_group_dev(&vdev->vdev);
-	kfree(vdev);
+	vfio_put_device(&vdev->vdev);
 	return 0;
 }
 
+static const struct vfio_device_ops vfio_fsl_mc_ops = {
+	.name		= "vfio-fsl-mc",
+	.init		= vfio_fsl_mc_init_dev,
+	.release	= vfio_fsl_mc_release_dev,
+	.open_device	= vfio_fsl_mc_open_device,
+	.close_device	= vfio_fsl_mc_close_device,
+	.ioctl		= vfio_fsl_mc_ioctl,
+	.read		= vfio_fsl_mc_read,
+	.write		= vfio_fsl_mc_write,
+	.mmap		= vfio_fsl_mc_mmap,
+};
+
 static struct fsl_mc_driver vfio_fsl_mc_driver = {
 	.probe		= vfio_fsl_mc_probe,
 	.remove		= vfio_fsl_mc_remove,
Signature · 1 line
2.21.3
author reviewer
#16 lore raw

PATCH v3 11/15 vfio/platform: Use the new device life cycle helpers

In reply to Kevin Tian

Move vfio_device_ops from platform core to platform drivers so device specific init/cleanup can be added.

Introduce two new helpers vfio_platform_init/release_common() for the use in driver @init/@release.

vfio_platform_probe/remove_common() will be deprecated.

---
 drivers/vfio/platform/vfio_platform.c         | 66 +++++++++++++++----
 drivers/vfio/platform/vfio_platform_common.c  | 53 ++++++++++++---
 drivers/vfio/platform/vfio_platform_private.h | 15 +++++
 3 files changed, 111 insertions(+), 23 deletions(-)
drivers/vfio/platform/vfio_platform.c +52 -14
index 04f40c5acfd6..82cedcebfd90 100644
@@ -7,6 +7,7 @@
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/vfio.h>
+#include <linux/pm_runtime.h>
 #include <linux/platform_device.h>
 
 #include "vfio_platform_private.h"
@@ -36,14 +37,11 @@ static int get_platform_irq(struct vfio_platform_device *vdev, int i)
 	return platform_get_irq_optional(pdev, i);
 }
 
-static int vfio_platform_probe(struct platform_device *pdev)
+static int vfio_platform_init_dev(struct vfio_device *core_vdev)
 {
-	struct vfio_platform_device *vdev;
-	int ret;
-
-	vdev = kzalloc(sizeof(*vdev), GFP_KERNEL);
-	if (!vdev)
-		return -ENOMEM;
+	struct vfio_platform_device *vdev =
+		container_of(core_vdev, struct vfio_platform_device, vdev);
+	struct platform_device *pdev = to_platform_device(core_vdev->dev);
 
 	vdev->opaque = (void *) pdev;
 	vdev->name = pdev->name;
@@ -52,24 +50,64 @@ static int vfio_platform_probe(struct platform_device *pdev)
 	vdev->get_irq = get_platform_irq;
 	vdev->reset_required = reset_required;
 
-	ret = vfio_platform_probe_common(vdev, &pdev->dev);
-	if (ret) {
-		kfree(vdev);
-		return ret;
-	}
+	return vfio_platform_init_common(vdev);
+}
+
+static const struct vfio_device_ops vfio_platform_ops;
+static int vfio_platform_probe(struct platform_device *pdev)
+{
+	struct vfio_platform_device *vdev;
+	int ret;
+
+	vdev = vfio_alloc_device(vfio_platform_device, vdev, &pdev->dev,
+				 &vfio_platform_ops);
+	if (IS_ERR(vdev))
+		return PTR_ERR(vdev);
+
+	ret = vfio_register_group_dev(&vdev->vdev);
+	if (ret)
+		goto out_put_vdev;
+
+	pm_runtime_enable(&pdev->dev);
 	dev_set_drvdata(&pdev->dev, vdev);
 	return 0;
+
+out_put_vdev:
+	vfio_put_device(&vdev->vdev);
+	return ret;
+}
+
+static void vfio_platform_release_dev(struct vfio_device *core_vdev)
+{
+	struct vfio_platform_device *vdev =
+		container_of(core_vdev, struct vfio_platform_device, vdev);
+
+	vfio_platform_release_common(vdev);
+	vfio_free_device(core_vdev);
 }
 
 static int vfio_platform_remove(struct platform_device *pdev)
 {
 	struct vfio_platform_device *vdev = dev_get_drvdata(&pdev->dev);
 
-	vfio_platform_remove_common(vdev);
-	kfree(vdev);
+	vfio_unregister_group_dev(&vdev->vdev);
+	pm_runtime_disable(vdev->device);
+	vfio_put_device(&vdev->vdev);
 	return 0;
 }
 
+static const struct vfio_device_ops vfio_platform_ops = {
+	.name		= "vfio-platform",
+	.init		= vfio_platform_init_dev,
+	.release	= vfio_platform_release_dev,
+	.open_device	= vfio_platform_open_device,
+	.close_device	= vfio_platform_close_device,
+	.ioctl		= vfio_platform_ioctl,
+	.read		= vfio_platform_read,
+	.write		= vfio_platform_write,
+	.mmap		= vfio_platform_mmap,
+};
+
 static struct platform_driver vfio_platform_driver = {
 	.probe		= vfio_platform_probe,
 	.remove		= vfio_platform_remove,
drivers/vfio/platform/vfio_platform_common.c +44 -9

121 lines not on this page · raw message

drivers/vfio/platform/vfio_platform_private.h +15 -0
index 691b43f4b2b2..a769d649fb97 100644
@@ -81,6 +81,21 @@ struct vfio_platform_reset_node {
 int vfio_platform_probe_common(struct vfio_platform_device *vdev,
 			       struct device *dev);
 void vfio_platform_remove_common(struct vfio_platform_device *vdev);
+int vfio_platform_init_common(struct vfio_platform_device *vdev);
+void vfio_platform_release_common(struct vfio_platform_device *vdev);
+
+int vfio_platform_open_device(struct vfio_device *core_vdev);
+void vfio_platform_close_device(struct vfio_device *core_vdev);
+long vfio_platform_ioctl(struct vfio_device *core_vdev,
+			 unsigned int cmd, unsigned long arg);
+ssize_t vfio_platform_read(struct vfio_device *core_vdev,
+			   char __user *buf, size_t count,
+			   loff_t *ppos);
+ssize_t vfio_platform_write(struct vfio_device *core_vdev,
+			    const char __user *buf,
+			    size_t count, loff_t *ppos);
+int vfio_platform_mmap(struct vfio_device *core_vdev,
+		       struct vm_area_struct *vma);
 
 int vfio_platform_irq_init(struct vfio_platform_device *vdev);
 void vfio_platform_irq_cleanup(struct vfio_platform_device *vdev);
Signature · 1 line
2.21.3
author reviewer
#17 lore raw

PATCH v3 12/15 vfio/amba: Use the new device life cycle helpers

In reply to Kevin Tian

Implement amba's own vfio_device_ops.

Remove vfio_platform_probe/remove_common() given no user now.

---
 drivers/vfio/platform/vfio_amba.c             | 72 ++++++++++++++-----
 drivers/vfio/platform/vfio_platform_common.c  | 60 ----------------
 drivers/vfio/platform/vfio_platform_private.h |  3 -
 3 files changed, 55 insertions(+), 80 deletions(-)
drivers/vfio/platform/vfio_amba.c +55 -17
index 1aaa4f721bd2..eaea63e5294c 100644
@@ -7,6 +7,7 @@
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/vfio.h>
+#include <linux/pm_runtime.h>
 #include <linux/amba/bus.h>
 
 #include "vfio_platform_private.h"
@@ -40,20 +41,16 @@ static int get_amba_irq(struct vfio_platform_device *vdev, int i)
 	return ret ? ret : -ENXIO;
 }
 
-static int vfio_amba_probe(struct amba_device *adev, const struct amba_id *id)
+static int vfio_amba_init_dev(struct vfio_device *core_vdev)
 {
-	struct vfio_platform_device *vdev;
+	struct vfio_platform_device *vdev =
+		container_of(core_vdev, struct vfio_platform_device, vdev);
+	struct amba_device *adev = to_amba_device(core_vdev->dev);
 	int ret;
 
-	vdev = kzalloc(sizeof(*vdev), GFP_KERNEL);
-	if (!vdev)
-		return -ENOMEM;
-
 	vdev->name = kasprintf(GFP_KERNEL, "vfio-amba-%08x", adev->periphid);
-	if (!vdev->name) {
-		kfree(vdev);
+	if (!vdev->name)
 		return -ENOMEM;
-	}
 
 	vdev->opaque = (void *) adev;
 	vdev->flags = VFIO_DEVICE_FLAGS_AMBA;
@@ -61,26 +58,67 @@ static int vfio_amba_probe(struct amba_device *adev, const struct amba_id *id)
 	vdev->get_irq = get_amba_irq;
 	vdev->reset_required = false;
 
-	ret = vfio_platform_probe_common(vdev, &adev->dev);
-	if (ret) {
+	ret = vfio_platform_init_common(vdev);
+	if (ret)
 		kfree(vdev->name);
-		kfree(vdev);
-		return ret;
-	}
+	return ret;
+}
+
+static const struct vfio_device_ops vfio_amba_ops;
+static int vfio_amba_probe(struct amba_device *adev, const struct amba_id *id)
+{
+	struct vfio_platform_device *vdev;
+	int ret;
+
+	vdev = vfio_alloc_device(vfio_platform_device, vdev, &adev->dev,
+				 &vfio_amba_ops);
+	if (IS_ERR(vdev))
+		return PTR_ERR(vdev);
 
+	ret = vfio_register_group_dev(&vdev->vdev);
+	if (ret)
+		goto out_put_vdev;
+
+	pm_runtime_enable(&adev->dev);
 	dev_set_drvdata(&adev->dev, vdev);
 	return 0;
+
+out_put_vdev:
+	vfio_put_device(&vdev->vdev);
+	return ret;
+}
+
+static void vfio_amba_release_dev(struct vfio_device *core_vdev)
+{
+	struct vfio_platform_device *vdev =
+		container_of(core_vdev, struct vfio_platform_device, vdev);
+
+	vfio_platform_release_common(vdev);
+	kfree(vdev->name);
+	vfio_free_device(core_vdev);
 }
 
 static void vfio_amba_remove(struct amba_device *adev)
 {
 	struct vfio_platform_device *vdev = dev_get_drvdata(&adev->dev);
 
-	vfio_platform_remove_common(vdev);
-	kfree(vdev->name);
-	kfree(vdev);
+	vfio_unregister_group_dev(&vdev->vdev);
+	pm_runtime_disable(vdev->device);
+	vfio_put_device(&vdev->vdev);
 }
 
+static const struct vfio_device_ops vfio_amba_ops = {
+	.name		= "vfio-amba",
+	.init		= vfio_amba_init_dev,
+	.release	= vfio_amba_release_dev,
+	.open_device	= vfio_platform_open_device,
+	.close_device	= vfio_platform_close_device,
+	.ioctl		= vfio_platform_ioctl,
+	.read		= vfio_platform_read,
+	.write		= vfio_platform_write,
+	.mmap		= vfio_platform_mmap,
+};
+
 static const struct amba_id pl330_ids[] = {
 	{ 0, 0 },
 };
drivers/vfio/platform/vfio_platform_common.c +0 -60

72 lines not on this page · raw message

drivers/vfio/platform/vfio_platform_private.h +0 -3
index a769d649fb97..8d8fab516849 100644
@@ -78,9 +78,6 @@ struct vfio_platform_reset_node {
 	vfio_platform_reset_fn_t of_reset;
 };
 
-int vfio_platform_probe_common(struct vfio_platform_device *vdev,
-			       struct device *dev);
-void vfio_platform_remove_common(struct vfio_platform_device *vdev);
 int vfio_platform_init_common(struct vfio_platform_device *vdev);
 void vfio_platform_release_common(struct vfio_platform_device *vdev);
 
Signature · 1 line
2.21.3
author reviewer
#18 lore raw

PATCH v3 13/15 vfio/ccw: Use the new device life cycle helpers

In reply to Kevin Tian

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 already 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.

---
 drivers/s390/cio/vfio_ccw_ops.c     | 52 +++++++++++++++++++++++++----
 drivers/s390/cio/vfio_ccw_private.h |  3 ++
 drivers/vfio/vfio_main.c            | 23 +++----------
 include/linux/vfio.h                |  3 --
 4 files changed, 53 insertions(+), 28 deletions(-)
drivers/s390/cio/vfio_ccw_ops.c +45 -7
index 4a806a2273b5..9f8486c0d3d3 100644
@@ -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,
drivers/s390/cio/vfio_ccw_private.h +3 -0
index cd24b7fada91..63d9202b29c7 100644
@@ -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);
drivers/vfio/vfio_main.c +5 -18
index adc1b697bb78..5bc221730d7b 100644
@@ -483,28 +483,13 @@ 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);
-
 /* 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);
+	vfio_release_device_set(device);
 
 	/*
 	 * kvfree() cannot be done here due to a life cycle mess in
@@ -562,7 +547,9 @@ 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);
@@ -574,7 +561,7 @@ int vfio_init_device(struct vfio_device *device, struct device *dev,
 	return 0;
 
 out_uninit:
-	vfio_uninit_group_dev(device);
+	vfio_release_device_set(device);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(vfio_init_device);
include/linux/vfio.h +0 -3
index f67cac700e6f..3cf857b1eec7 100644
@@ -184,9 +184,6 @@ 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);
Signature · 1 line
2.21.3
author reviewer
#19 lore raw

PATCH v3 14/15 vfio: Rename vfio_device_put() and vfio_device_try_get()

In reply to Kevin Tian

With the addition of vfio_put_device() now the names become confusing.

vfio_put_device() is clear from object life cycle p.o.v given kref.

vfio_device_put()/vfio_device_try_get() are helpers for tracking users on a registered device.

Now rename them:

- vfio_device_put() -> vfio_device_put_registration()
- vfio_device_try_get() -> vfio_device_try_get_registration()

---
 drivers/vfio/vfio_main.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)
drivers/vfio/vfio_main.c +9 -8
index 5bc221730d7b..37cbd10f3faf 100644
@@ -453,13 +453,13 @@ static void vfio_group_get(struct vfio_group *group)
  * Device objects - create, release, get, put, search
  */
 /* Device reference always implies a group reference */
-static void vfio_device_put(struct vfio_device *device)
+static void vfio_device_put_registration(struct vfio_device *device)
 {
 	if (refcount_dec_and_test(&device->refcount))
 		complete(&device->comp);
 }
 
-static bool vfio_device_try_get(struct vfio_device *device)
+static bool vfio_device_try_get_registration(struct vfio_device *device)
 {
 	return refcount_inc_not_zero(&device->refcount);
 }
@@ -471,7 +471,8 @@ static struct vfio_device *vfio_group_get_device(struct vfio_group *group,
 
 	mutex_lock(&group->device_lock);
 	list_for_each_entry(device, &group->device_list, group_next) {
-		if (device->dev == dev && vfio_device_try_get(device)) {
+		if (device->dev == dev &&
+		    vfio_device_try_get_registration(device)) {
 			mutex_unlock(&group->device_lock);
 			return device;
 		}
@@ -673,7 +674,7 @@ static int __vfio_register_dev(struct vfio_device *device,
 	if (existing_device) {
 		dev_WARN(device->dev, "Device already exists on group %d\n",
 			 iommu_group_id(group->iommu_group));
-		vfio_device_put(existing_device);
+		vfio_device_put_registration(existing_device);
 		if (group->type == VFIO_NO_IOMMU ||
 		    group->type == VFIO_EMULATED_IOMMU)
 			iommu_group_remove_device(device->dev);
@@ -731,7 +732,7 @@ static struct vfio_device *vfio_device_get_from_name(struct vfio_group *group,
 			ret = !strcmp(dev_name(it->dev), buf);
 		}
 
-		if (ret && vfio_device_try_get(it)) {
+		if (ret && vfio_device_try_get_registration(it)) {
 			device = it;
 			break;
 		}
@@ -751,7 +752,7 @@ void vfio_unregister_group_dev(struct vfio_device *device)
 	bool interrupted = false;
 	long rc;
 
-	vfio_device_put(device);
+	vfio_device_put_registration(device);
 	rc = try_wait_for_completion(&device->comp);
 	while (rc <= 0) {
 		if (device->ops->request)
@@ -1311,7 +1312,7 @@ static int vfio_group_ioctl_get_device_fd(struct vfio_group *group,
 err_put_fdno:
 	put_unused_fd(fdno);
 err_put_device:
-	vfio_device_put(device);
+	vfio_device_put_registration(device);
 	return ret;
 }
 
@@ -1493,7 +1494,7 @@ static int vfio_device_fops_release(struct inode *inode, struct file *filep)
 
 	vfio_device_unassign_container(device);
 
-	vfio_device_put(device);
+	vfio_device_put_registration(device);
 
 	return 0;
 }
Signature · 1 line
2.21.3
author reviewer
#20 lore raw

PATCH v3 15/15 vfio: Add struct device to vfio_device

In reply to Kevin Tian

From: Yi Liu <[email protected]>

and replace kref. With it a 'vfio-dev/vfioX' node is created under the sysfs path of the parent, indicating the device is bound to a vfio driver, e.g.:

/sys/devices/pci0000\:6f/0000\:6f\:01.0/vfio-dev/vfio0

It is also a preparatory step toward adding cdev for supporting future device-oriented uAPI.

Add Documentation/ABI/testing/sysfs-devices-vfio-dev.

Also take this chance to rename chardev 'vfio' to 'vfio-group' in /proc/devices.

---
 .../ABI/testing/sysfs-devices-vfio-dev        |  8 +++
 MAINTAINERS                                   |  1 +
 drivers/vfio/vfio_main.c                      | 67 +++++++++++++++----
 include/linux/vfio.h                          |  6 +-
 4 files changed, 67 insertions(+), 15 deletions(-)
 create mode 100644 Documentation/ABI/testing/sysfs-devices-vfio-dev
Documentation/ABI/testing/sysfs-devices-vfio-dev new file +8 -0
new file mode 100644
index 000000000000..e21424fd9666
@@ -0,0 +1,8 @@
+What:		 /sys/.../<device>/vfio-dev/vfioX/
+Date:		 September 2022
+Contact:	 Yi Liu <[email protected]>
+Description:
+		 This directory is created when the device is bound to a
+		 vfio driver. The layout under this directory matches what
+		 exists for a standard 'struct device'. 'X' is a unique
+		 index marking this device in vfio.
MAINTAINERS +1 -0
index d30f26e07cd3..02c8f11b1c17 100644
@@ -21312,6 +21312,7 @@ R:	Cornelia Huck <[email protected]>
 L:	[email protected]
 S:	Maintained
 T:	git git://github.com/awilliam/linux-vfio.git
+F:	Documentation/ABI/testing/sysfs-devices-vfio-dev
 F:	Documentation/driver-api/vfio.rst
 F:	drivers/vfio/
 F:	include/linux/vfio.h
drivers/vfio/vfio_main.c +55 -12

166 lines not on this page · raw message

include/linux/vfio.h +3 -3
index 3cf857b1eec7..ee399a768070 100644
@@ -47,7 +47,8 @@ struct vfio_device {
 	struct kvm *kvm;
 
 	/* Members below here are private, not for driver use */
-	struct kref kref;	/* object life cycle */
+	unsigned int index;
+	struct device device;	/* device.kref covers object life circle */
 	refcount_t refcount;	/* user count on registered device*/
 	unsigned int open_count;
 	struct completion comp;
@@ -178,10 +179,9 @@ struct vfio_device *_vfio_alloc_device(size_t size, struct device *dev,
 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);
+	put_device(&device->device);
 }
 
 int vfio_register_group_dev(struct vfio_device *device);
Signature · 1 line
2.21.3
#21 lore raw

Re: [PATCH v3 15/15] vfio: Add struct device to vfio_device

Quoted 8 lines · On Fri, 9 Sep 2022 18:22:47 +0800 Kevin Tian <[email protected]> wrote:

From: Yi Liu <[email protected]>

and replace kref. With it a 'vfio-dev/vfioX' node is created under the sysfs path of the parent, indicating the device is bound to a vfio driver, e.g.:

/sys/devices/pci0000\:6f/0000\:6f\:01.0/vfio-dev/vfio0

Quoted 7 lines

It is also a preparatory step toward adding cdev for supporting future device-oriented uAPI.

Add Documentation/ABI/testing/sysfs-devices-vfio-dev.

Also take this chance to rename chardev 'vfio' to 'vfio-group' in /proc/devices.

What's the risk/reward here, is this just more aesthetically pleasing symmetry vs 'vfio-dev'? The char major number to name association in /proc/devices seems pretty obscure, but what due diligence have we done to make sure this doesn't break anyone? Thanks,

Alex

author reviewer
#22 lore raw

RE: [PATCH v3 15/15] vfio: Add struct device to vfio_device

Quoted 19 lines

From: Alex Williamson <[email protected]>
Sent: Wednesday, September 21, 2022 4:27 AM

Quoted 13 lines · On Fri, 9 Sep 2022 18:22:47 +0800 Kevin Tian <[email protected]> wrote:

From: Yi Liu <[email protected]>

and replace kref. With it a 'vfio-dev/vfioX' node is created under the sysfs path of the parent, indicating the device is bound to a vfio driver, e.g.:

/sys/devices/pci0000\:6f/0000\:6f\:01.0/vfio-dev/vfio0

It is also a preparatory step toward adding cdev for supporting future device-oriented uAPI.

Add Documentation/ABI/testing/sysfs-devices-vfio-dev.

Quoted 8 lines
Quoted 2 lines

Also take this chance to rename chardev 'vfio' to 'vfio-group' in /proc/devices.

What's the risk/reward here, is this just more aesthetically pleasing symmetry vs 'vfio-dev'? The char major number to name association in /proc/devices seems pretty obscure, but what due diligence have we done to make sure this doesn't break anyone? Thanks,

I'm not sure whether the content of /proc/devices is considered as ABI.

@Jason?

But to be safe I can remove this change in next version. If it's the right thing to do such change after discussion then it can be done in a separate patch after.

#23 lore raw

Re: [PATCH v3 15/15] vfio: Add struct device to vfio_device

Quoted 22 lines · Kevin
Quoted 22 lines

From: Alex Williamson <[email protected]>
Sent: Wednesday, September 21, 2022 4:27 AM

Quoted 15 lines · On Fri, 9 Sep 2022 18:22:47 +0800 Kevin Tian <[email protected]> wrote:

From: Yi Liu <[email protected]>

and replace kref. With it a 'vfio-dev/vfioX' node is created under the sysfs path of the parent, indicating the device is bound to a vfio driver, e.g.:

/sys/devices/pci0000\:6f/0000\:6f\:01.0/vfio-dev/vfio0

It is also a preparatory step toward adding cdev for supporting future device-oriented uAPI.

Add Documentation/ABI/testing/sysfs-devices-vfio-dev.

Also take this chance to rename chardev 'vfio' to 'vfio-group' in /proc/devices.

Quoted 8 lines
Quoted 4 lines

What's the risk/reward here, is this just more aesthetically pleasing symmetry vs 'vfio-dev'? The char major number to name association in /proc/devices seems pretty obscure, but what due diligence have we done to make sure this doesn't break anyone? Thanks,

I'm not sure whether the content of /proc/devices is considered as ABI.

@Jason?

Ah, I've forgotten why we got here - didn't we have a naming conflict with the new stuff that is being introduced?

ABI wise it is not a problem unless there is a real user, I'm not aware of anything scanning /proc, that has been obsoleted by sysfs a long time ago.

Jason

author reviewer
#24 lore raw

RE: [PATCH v3 15/15] vfio: Add struct device to vfio_device

Quoted 30 lines

From: Jason Gunthorpe <[email protected]>
Sent: Thursday, September 22, 2022 12:10 AM

Quoted 17 lines · Kevin
Quoted 17 lines

From: Alex Williamson <[email protected]>
Sent: Wednesday, September 21, 2022 4:27 AM

Quoted 11 lines · On Fri, 9 Sep 2022 18:22:47 +0800 Kevin Tian <[email protected]> wrote:

From: Yi Liu <[email protected]>

and replace kref. With it a 'vfio-dev/vfioX' node is created under the sysfs path of the parent, indicating the device is bound to a vfio driver, e.g.:

/sys/devices/pci0000\:6f/0000\:6f\:01.0/vfio-dev/vfio0

It is also a preparatory step toward adding cdev for supporting future device-oriented uAPI.

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

Add Documentation/ABI/testing/sysfs-devices-vfio-dev.

Also take this chance to rename chardev 'vfio' to 'vfio-group' in /proc/devices.

What's the risk/reward here, is this just more aesthetically pleasing symmetry vs 'vfio-dev'? The char major number to name association in /proc/devices seems pretty obscure, but what due diligence have we

done

Quoted 8 lines
Quoted 5 lines
Quoted 1 line

to make sure this doesn't break anyone? Thanks,

I'm not sure whether the content of /proc/devices is considered as ABI.

@Jason?

Ah, I've forgotten why we got here - didn't we have a naming conflict with the new stuff that is being introduced?

No, we don't have. There is no new char dev introduced in this series.

Later when device cdev is added a new device major will be allocated for 'vfio-dev'. It's at that time renaming existing 'vfio' to 'vfio-group' is probably clearer, which is what I understood from your earlier suggestion.

Quoted 5 lines

ABI wise it is not a problem unless there is a real user, I'm not aware of anything scanning /proc, that has been obsoleted by sysfs a long time ago.

This is a good news.

#25 lore raw

[Intel-gfx] ✗ Fi.CI.BUILD: failure for Tidy up vfio_device life cycle (rev6)

In reply to Kevin Tian
== Series Details ==

== Series Details ==

Series: Tidy up vfio_device life cycle (rev6)
URL : https://patchwork.freedesktop.org/series/107838/
State : failure

== Summary ==

Error: patch https://patchwork.freedesktop.org/api/1.0/series/107838/revisions/6/mbox/ not applied
Applying: vfio: Add helpers for unifying vfio_device life cycle
Applying: vfio/pci: Use the new device life cycle helpers
Applying: vfio/mlx5: Use the new device life cycle helpers
error: sha1 information is lacking or useless (drivers/vfio/pci/mlx5/main.c).
error: could not build fake ancestor
hint: Use 'git am --show-current-patch=diff' to see the failed patch
Patch failed at 0003 vfio/mlx5: Use the new device life cycle helpers
When you have resolved this problem, run "git am --continue".
If you prefer to skip this patch, run "git am --skip" instead.
To restore the original branch and stop patching, run "git am --abort".