vfio-iommufd: Support iommufd for physical VFIO devices

This creates the iommufd_device for the physical VFIO drivers. These are
all the drivers that are calling vfio_register_group_dev() and expect the
type1 code to setup a real iommu_domain against their parent struct
device.

The design gives the driver a choice in how it gets connected to iommufd
by providing bind_iommufd/unbind_iommufd/attach_ioas callbacks to
implement as required. The core code provides three default callbacks for
physical mode using a real iommu_domain. This is suitable for drivers
using vfio_register_group_dev()

Link: https://lore.kernel.org/r/6-v4-42cd2eb0e3eb+335a-vfio_iommufd_jgg@nvidia.comReviewed-by: Kevin Tian <kevin.tian@intel.com>
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Tested-by: Alex Williamson <alex.williamson@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Tested-by: Yi Liu <yi.l.liu@intel.com>
Tested-by: Lixiao Yang <lixiao.yang@intel.com>
Tested-by: Matthew Rosato <mjrosato@linux.ibm.com>
Tested-by: Yu He <yu.he@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

drivers/vfio/Makefile

@@ -6,6 +6,7 @@ obj-$(CONFIG_VFIO) += vfio.o
 vfio-y += vfio_main.o \
 	  iova_bitmap.o \
 	  container.o
+vfio-$(CONFIG_IOMMUFD) += iommufd.o
 
 obj-$(CONFIG_VFIO_VIRQFD) += vfio_virqfd.o
 obj-$(CONFIG_VFIO_IOMMU_TYPE1) += vfio_iommu_type1.o

drivers/vfio/fsl-mc/vfio_fsl_mc.c

@@ -592,6 +592,9 @@ static const struct vfio_device_ops vfio_fsl_mc_ops = {
 	.read		= vfio_fsl_mc_read,
 	.write		= vfio_fsl_mc_write,
 	.mmap		= vfio_fsl_mc_mmap,
+	.bind_iommufd	= vfio_iommufd_physical_bind,
+	.unbind_iommufd	= vfio_iommufd_physical_unbind,
+	.attach_ioas	= vfio_iommufd_physical_attach_ioas,
 };
 
 static struct fsl_mc_driver vfio_fsl_mc_driver = {

drivers/vfio/iommufd.c

+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2021-2022, NVIDIA CORPORATION & AFFILIATES
+ */
+#include <linux/vfio.h>
+#include <linux/iommufd.h>
+
+#include "vfio.h"
+
+MODULE_IMPORT_NS(IOMMUFD);
+MODULE_IMPORT_NS(IOMMUFD_VFIO);
+
+int vfio_iommufd_bind(struct vfio_device *vdev, struct iommufd_ctx *ictx)
+{
+	u32 ioas_id;
+	u32 device_id;
+	int ret;
+
+	lockdep_assert_held(&vdev->dev_set->lock);
+
+	/*
+	 * If the driver doesn't provide this op then it means the device does
+	 * not do DMA at all. So nothing to do.
+	 */
+	if (!vdev->ops->bind_iommufd)
+		return 0;
+
+	ret = vdev->ops->bind_iommufd(vdev, ictx, &device_id);
+	if (ret)
+		return ret;
+
+	ret = iommufd_vfio_compat_ioas_id(ictx, &ioas_id);
+	if (ret)
+		goto err_unbind;
+	ret = vdev->ops->attach_ioas(vdev, &ioas_id);
+	if (ret)
+		goto err_unbind;
+
+	/*
+	 * The legacy path has no way to return the device id or the selected
+	 * pt_id
+	 */
+	return 0;
+
+err_unbind:
+	if (vdev->ops->unbind_iommufd)
+		vdev->ops->unbind_iommufd(vdev);
+	return ret;
+}
+
+void vfio_iommufd_unbind(struct vfio_device *vdev)
+{
+	lockdep_assert_held(&vdev->dev_set->lock);
+
+	if (vdev->ops->unbind_iommufd)
+		vdev->ops->unbind_iommufd(vdev);
+}
+
+/*
+ * The physical standard ops mean that the iommufd_device is bound to the
+ * physical device vdev->dev that was provided to vfio_init_group_dev(). Drivers
+ * using this ops set should call vfio_register_group_dev()
+ */
+int vfio_iommufd_physical_bind(struct vfio_device *vdev,
+			       struct iommufd_ctx *ictx, u32 *out_device_id)
+{
+	struct iommufd_device *idev;
+
+	idev = iommufd_device_bind(ictx, vdev->dev, out_device_id);
+	if (IS_ERR(idev))
+		return PTR_ERR(idev);
+	vdev->iommufd_device = idev;
+	return 0;
+}
+EXPORT_SYMBOL_GPL(vfio_iommufd_physical_bind);
+
+void vfio_iommufd_physical_unbind(struct vfio_device *vdev)
+{
+	lockdep_assert_held(&vdev->dev_set->lock);
+
+	if (vdev->iommufd_attached) {
+		iommufd_device_detach(vdev->iommufd_device);
+		vdev->iommufd_attached = false;
+	}
+	iommufd_device_unbind(vdev->iommufd_device);
+	vdev->iommufd_device = NULL;
+}
+EXPORT_SYMBOL_GPL(vfio_iommufd_physical_unbind);
+
+int vfio_iommufd_physical_attach_ioas(struct vfio_device *vdev, u32 *pt_id)
+{
+	int rc;
+
+	rc = iommufd_device_attach(vdev->iommufd_device, pt_id);
+	if (rc)
+		return rc;
+	vdev->iommufd_attached = true;
+	return 0;
+}
+EXPORT_SYMBOL_GPL(vfio_iommufd_physical_attach_ioas);

drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c

@@ -1246,6 +1246,9 @@ static const struct vfio_device_ops hisi_acc_vfio_pci_migrn_ops = {
 	.mmap = hisi_acc_vfio_pci_mmap,
 	.request = vfio_pci_core_request,
 	.match = vfio_pci_core_match,
+	.bind_iommufd = vfio_iommufd_physical_bind,
+	.unbind_iommufd = vfio_iommufd_physical_unbind,
+	.attach_ioas = vfio_iommufd_physical_attach_ioas,
 };
 
 static const struct vfio_device_ops hisi_acc_vfio_pci_ops = {
@@ -1261,6 +1264,9 @@ static const struct vfio_device_ops hisi_acc_vfio_pci_ops = {
 	.mmap = vfio_pci_core_mmap,
 	.request = vfio_pci_core_request,
 	.match = vfio_pci_core_match,
+	.bind_iommufd = vfio_iommufd_physical_bind,
+	.unbind_iommufd = vfio_iommufd_physical_unbind,
+	.attach_ioas = vfio_iommufd_physical_attach_ioas,
 };
 
 static int hisi_acc_vfio_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)

drivers/vfio/pci/mlx5/main.c

@@ -623,6 +623,9 @@ static const struct vfio_device_ops mlx5vf_pci_ops = {
 	.mmap = vfio_pci_core_mmap,
 	.request = vfio_pci_core_request,
 	.match = vfio_pci_core_match,
+	.bind_iommufd = vfio_iommufd_physical_bind,
+	.unbind_iommufd = vfio_iommufd_physical_unbind,
+	.attach_ioas = vfio_iommufd_physical_attach_ioas,
 };
 
 static int mlx5vf_pci_probe(struct pci_dev *pdev,

drivers/vfio/pci/vfio_pci.c

@@ -138,6 +138,9 @@ static const struct vfio_device_ops vfio_pci_ops = {
 	.mmap		= vfio_pci_core_mmap,
 	.request	= vfio_pci_core_request,
 	.match		= vfio_pci_core_match,
+	.bind_iommufd	= vfio_iommufd_physical_bind,
+	.unbind_iommufd	= vfio_iommufd_physical_unbind,
+	.attach_ioas	= vfio_iommufd_physical_attach_ioas,
 };
 
 static int vfio_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)

drivers/vfio/platform/vfio_amba.c

@@ -117,6 +117,9 @@ static const struct vfio_device_ops vfio_amba_ops = {
 	.read		= vfio_platform_read,
 	.write		= vfio_platform_write,
 	.mmap		= vfio_platform_mmap,
+	.bind_iommufd	= vfio_iommufd_physical_bind,
+	.unbind_iommufd	= vfio_iommufd_physical_unbind,
+	.attach_ioas	= vfio_iommufd_physical_attach_ioas,
 };
 
 static const struct amba_id pl330_ids[] = {

drivers/vfio/platform/vfio_platform.c

@@ -106,6 +106,9 @@ static const struct vfio_device_ops vfio_platform_ops = {
 	.read		= vfio_platform_read,
 	.write		= vfio_platform_write,
 	.mmap		= vfio_platform_mmap,
+	.bind_iommufd	= vfio_iommufd_physical_bind,
+	.unbind_iommufd	= vfio_iommufd_physical_unbind,
+	.attach_ioas	= vfio_iommufd_physical_attach_ioas,
 };
 
 static struct platform_driver vfio_platform_driver = {

drivers/vfio/vfio.h

@@ -124,6 +124,21 @@ void vfio_device_container_unregister(struct vfio_device *device);
 int __init vfio_container_init(void);
 void vfio_container_cleanup(void);
 
+#if IS_ENABLED(CONFIG_IOMMUFD)
+int vfio_iommufd_bind(struct vfio_device *device, struct iommufd_ctx *ictx);
+void vfio_iommufd_unbind(struct vfio_device *device);
+#else
+static inline int vfio_iommufd_bind(struct vfio_device *device,
+				    struct iommufd_ctx *ictx)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline void vfio_iommufd_unbind(struct vfio_device *device)
+{
+}
+#endif
+
 #ifdef CONFIG_VFIO_NOIOMMU
 extern bool vfio_noiommu __read_mostly;
 #else

drivers/vfio/vfio_main.c

@@ -525,6 +525,11 @@ static int __vfio_register_dev(struct vfio_device *device,
 	if (IS_ERR(group))
 		return PTR_ERR(group);
 
+	if (WARN_ON(device->ops->bind_iommufd &&
+		    (!device->ops->unbind_iommufd ||
+		     !device->ops->attach_ioas)))
+		return -EINVAL;
+
 	/*
 	 * If the driver doesn't specify a set then the device is added to a
 	 * singleton set just for itself.
@@ -794,6 +799,10 @@ static int vfio_device_first_open(struct vfio_device *device)
 		ret = vfio_group_use_container(device->group);
 		if (ret)
 			goto err_module_put;
+	} else if (device->group->iommufd) {
+		ret = vfio_iommufd_bind(device, device->group->iommufd);
+		if (ret)
+			goto err_module_put;
 	}
 
 	device->kvm = device->group->kvm;
@@ -811,6 +820,8 @@ static int vfio_device_first_open(struct vfio_device *device)
 	device->kvm = NULL;
 	if (device->group->container)
 		vfio_group_unuse_container(device->group);
+	else if (device->group->iommufd)
+		vfio_iommufd_unbind(device);
 err_module_put:
 	mutex_unlock(&device->group->group_lock);
 	module_put(device->dev->driver->owner);
@@ -829,6 +840,8 @@ static void vfio_device_last_close(struct vfio_device *device)
 	device->kvm = NULL;
 	if (device->group->container)
 		vfio_group_unuse_container(device->group);
+	else if (device->group->iommufd)
+		vfio_iommufd_unbind(device);
 	mutex_unlock(&device->group->group_lock);
 	module_put(device->dev->driver->owner);
 }
@@ -1936,8 +1949,6 @@ static void __exit vfio_cleanup(void)
 module_init(vfio_init);
 module_exit(vfio_cleanup);
 
-MODULE_IMPORT_NS(IOMMUFD);
-MODULE_IMPORT_NS(IOMMUFD_VFIO);
 MODULE_VERSION(DRIVER_VERSION);
 MODULE_LICENSE("GPL v2");
 MODULE_AUTHOR(DRIVER_AUTHOR);

include/linux/vfio.h

@@ -17,6 +17,8 @@
 #include <linux/iova_bitmap.h>
 
 struct kvm;
+struct iommufd_ctx;
+struct iommufd_device;
 
 /*
  * VFIO devices can be placed in a set, this allows all devices to share this
@@ -54,6 +56,10 @@ struct vfio_device {
 	struct completion comp;
 	struct list_head group_next;
 	struct list_head iommu_entry;
+#if IS_ENABLED(CONFIG_IOMMUFD)
+	struct iommufd_device *iommufd_device;
+	bool iommufd_attached;
+#endif
 };
 
 /**
@@ -80,6 +86,10 @@ struct vfio_device_ops {
 	char	*name;
 	int	(*init)(struct vfio_device *vdev);
 	void	(*release)(struct vfio_device *vdev);
+	int	(*bind_iommufd)(struct vfio_device *vdev,
+				struct iommufd_ctx *ictx, u32 *out_device_id);
+	void	(*unbind_iommufd)(struct vfio_device *vdev);
+	int	(*attach_ioas)(struct vfio_device *vdev, u32 *pt_id);
 	int	(*open_device)(struct vfio_device *vdev);
 	void	(*close_device)(struct vfio_device *vdev);
 	ssize_t	(*read)(struct vfio_device *vdev, char __user *buf,
@@ -96,6 +106,21 @@ struct vfio_device_ops {
 				  void __user *arg, size_t argsz);
 };
 
+#if IS_ENABLED(CONFIG_IOMMUFD)
+int vfio_iommufd_physical_bind(struct vfio_device *vdev,
+			       struct iommufd_ctx *ictx, u32 *out_device_id);
+void vfio_iommufd_physical_unbind(struct vfio_device *vdev);
+int vfio_iommufd_physical_attach_ioas(struct vfio_device *vdev, u32 *pt_id);
+#else
+#define vfio_iommufd_physical_bind                                      \
+	((int (*)(struct vfio_device *vdev, struct iommufd_ctx *ictx,   \
+		  u32 *out_device_id)) NULL)
+#define vfio_iommufd_physical_unbind \
+	((void (*)(struct vfio_device *vdev)) NULL)
+#define vfio_iommufd_physical_attach_ioas \
+	((int (*)(struct vfio_device *vdev, u32 *pt_id)) NULL)
+#endif
+
 /**
  * @migration_set_state: Optional callback to change the migration state for
  *         devices that support migration. It's mandatory for