[v2,3/4] iommufd: Add IOMMU_DEVICE_GET_HW_INFO

  Under nested IOMMU translation, userspace owns the stage-1 translation
table (e.g. the stage-1 page table of Intel VT-d or the context table
of ARM SMMUv3, and etc.). Stage-1 translation tables are vendor specific,
and needs to be compatiable with the underlying IOMMU hardware. Hence,
userspace should know the IOMMU hardware capability before creating and
configuring the stage-1 translation table to kernel.

This adds IOMMU_DEVICE_GET_HW_INFO to query the IOMMU hardware information
for a given device. The returned data is vendor specific, userspace needs
to decode it with the structure mapped by the @out_data_type field.

As only physical devices have IOMMU hardware, so this will return error
if the given device is not a physical device.

Co-developed-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
---
 drivers/iommu/iommufd/device.c          | 74 +++++++++++++++++++++++++
 drivers/iommu/iommufd/iommufd_private.h |  1 +
 drivers/iommu/iommufd/main.c            |  3 +
 include/uapi/linux/iommufd.h            | 40 +++++++++++++
 4 files changed, 118 insertions(+)
  

On 2023/3/9 15:53, Yi Liu wrote:
> Under nested IOMMU translation, userspace owns the stage-1 translation
> table (e.g. the stage-1 page table of Intel VT-d or the context table
> of ARM SMMUv3, and etc.). Stage-1 translation tables are vendor specific,
> and needs to be compatiable with the underlying IOMMU hardware. Hence,
> userspace should know the IOMMU hardware capability before creating and
> configuring the stage-1 translation table to kernel.
> 
> This adds IOMMU_DEVICE_GET_HW_INFO to query the IOMMU hardware information
> for a given device. The returned data is vendor specific, userspace needs
> to decode it with the structure mapped by the @out_data_type field.
> 
> As only physical devices have IOMMU hardware, so this will return error
> if the given device is not a physical device.
> 
> Co-developed-by: Nicolin Chen <nicolinc@nvidia.com>
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> Signed-off-by: Yi Liu <yi.l.liu@intel.com>
> ---
>   drivers/iommu/iommufd/device.c          | 74 +++++++++++++++++++++++++
>   drivers/iommu/iommufd/iommufd_private.h |  1 +
>   drivers/iommu/iommufd/main.c            |  3 +
>   include/uapi/linux/iommufd.h            | 40 +++++++++++++
>   4 files changed, 118 insertions(+)
> 
> diff --git a/drivers/iommu/iommufd/device.c b/drivers/iommu/iommufd/device.c
> index c10e02f6a0be..6948539488a5 100644
> --- a/drivers/iommu/iommufd/device.c
> +++ b/drivers/iommu/iommufd/device.c
> @@ -257,6 +257,80 @@ struct iommufd_ctx *iommufd_device_to_ictx(struct iommufd_device *idev)
>   }
>   EXPORT_SYMBOL_NS_GPL(iommufd_device_to_ictx, IOMMUFD);
>   
> +static int iommufd_zero_fill_user(u64 ptr, int bytes)
> +{
> +	int index = 0;
> +
> +	for (; index < bytes; index++) {
> +		if (put_user(0, (uint8_t __user *)(ptr + index)))
> +			return -EFAULT;
> +	}
> +	return 0;
> +}
> +
> +int iommufd_device_get_hw_info(struct iommufd_ucmd *ucmd)
> +{
> +	struct iommu_hw_info *cmd = ucmd->cmd;
> +	struct iommufd_device *idev;
> +	const struct iommu_ops *ops;
> +	void *data;
> +	unsigned int length, data_len;
> +	int rc;

Reverse christmas tree order for declarations.

> +
> +	if (cmd->flags || cmd->__reserved || !cmd->data_len)
> +		return -EOPNOTSUPP;
> +
> +	idev = iommufd_get_device(ucmd, cmd->dev_id);
> +	if (IS_ERR(idev))
> +		return PTR_ERR(idev);
> +
> +	ops = dev_iommu_ops(idev->dev);
> +	if (!ops || !ops->hw_info) {

dev_iommu_ops() will never return a NULL.

Need below check

	dev->iommu && dev->iommu->iommu_dev

before dev_iommu_ops(). Perhaps something like below?

	if (!dev->iommu || !dev->iommu->iommu_dev)
		return -EINVAL;

	ops = dev_iommu_ops(idev->dev);
	if (!ops->hw_info)
		return -ENODEV;

> +		rc = -EOPNOTSUPP;
> +		goto out_put;
> +	}
> +
> +	/* driver has hw_info callback should have a unique driver_type */
> +	if (WARN_ON(ops->driver_type == IOMMU_HW_INFO_TYPE_DEFAULT)) {

If so, perhaps IOMMU_HW_INFO_TYPE_INVALID is more readable?

I'm not sure if a calltrace is really necessary here. If not, perhaps,

	if (ops->driver_type == IOMMU_HW_INFO_TYPE_DEFAULT) {
		pr_warn_ratelimited("iommu driver set an invalid type\n");
		rc = -ENODEV;
		goto out_put;
	}

> +		rc = -EOPNOTSUPP;
> +		goto out_put;
> +	}
> +
> +	data = ops->hw_info(idev->dev, &data_len);
> +	if (IS_ERR(data)) {
> +		rc = PTR_ERR(data);
> +		goto out_put;
> +	}
> +
> +	length = min(cmd->data_len, data_len);
> +	if (copy_to_user(u64_to_user_ptr(cmd->data_ptr), data, length)) {
> +		rc = -EFAULT;
> +		goto out_free_data;
> +	}
> +
> +	/*
> +	 * Zero the trailing bytes if the user buffer is bigger than the
> +	 * data size kernel actually has.
> +	 */
> +	if (length < cmd->data_len) {
> +		rc = iommufd_zero_fill_user(cmd->data_ptr + length,
> +					    cmd->data_len - length);
> +		if (rc)
> +			goto out_free_data;
> +	}
> +
> +	cmd->out_data_type = ops->driver_type;
> +	cmd->data_len = length;
> +
> +	rc = iommufd_ucmd_respond(ucmd, sizeof(*cmd));
> +
> +out_free_data:
> +	kfree(data);
> +out_put:
> +	iommufd_put_object(&idev->obj);
> +	return rc;
> +}
> +
>   static int iommufd_group_setup_msi(struct iommufd_group *igroup,
>   				   struct iommufd_hw_pagetable *hwpt)
>   {
> diff --git a/drivers/iommu/iommufd/iommufd_private.h b/drivers/iommu/iommufd/iommufd_private.h
> index b18f843ad6a4..05b5ad66f716 100644
> --- a/drivers/iommu/iommufd/iommufd_private.h
> +++ b/drivers/iommu/iommufd/iommufd_private.h
> @@ -307,6 +307,7 @@ iommufd_get_device(struct iommufd_ucmd *ucmd, u32 id)
>   }
>   
>   void iommufd_device_destroy(struct iommufd_object *obj);
> +int iommufd_device_get_hw_info(struct iommufd_ucmd *ucmd);
>   
>   struct iommufd_access {
>   	struct iommufd_object obj;
> diff --git a/drivers/iommu/iommufd/main.c b/drivers/iommu/iommufd/main.c
> index 694da191e4b1..f079c0bda46b 100644
> --- a/drivers/iommu/iommufd/main.c
> +++ b/drivers/iommu/iommufd/main.c
> @@ -262,6 +262,7 @@ static int iommufd_option(struct iommufd_ucmd *ucmd)
>   union ucmd_buffer {
>   	struct iommu_destroy destroy;
>   	struct iommu_hwpt_alloc hwpt;
> +	struct iommu_hw_info info;
>   	struct iommu_ioas_alloc alloc;
>   	struct iommu_ioas_allow_iovas allow_iovas;
>   	struct iommu_ioas_copy ioas_copy;
> @@ -295,6 +296,8 @@ static const struct iommufd_ioctl_op iommufd_ioctl_ops[] = {
>   	IOCTL_OP(IOMMU_DESTROY, iommufd_destroy, struct iommu_destroy, id),
>   	IOCTL_OP(IOMMU_HWPT_ALLOC, iommufd_hwpt_alloc, struct iommu_hwpt_alloc,
>   		 __reserved),
> +	IOCTL_OP(IOMMU_DEVICE_GET_HW_INFO, iommufd_device_get_hw_info,
> +		 struct iommu_hw_info, __reserved),
>   	IOCTL_OP(IOMMU_IOAS_ALLOC, iommufd_ioas_alloc_ioctl,
>   		 struct iommu_ioas_alloc, out_ioas_id),
>   	IOCTL_OP(IOMMU_IOAS_ALLOW_IOVAS, iommufd_ioas_allow_iovas,
> diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h
> index 955cbef640da..4ac525897b82 100644
> --- a/include/uapi/linux/iommufd.h
> +++ b/include/uapi/linux/iommufd.h
> @@ -46,6 +46,7 @@ enum {
>   	IOMMUFD_CMD_OPTION,
>   	IOMMUFD_CMD_VFIO_IOAS,
>   	IOMMUFD_CMD_HWPT_ALLOC,
> +	IOMMUFD_CMD_DEVICE_GET_HW_INFO,
>   };
>   
>   /**
> @@ -377,4 +378,43 @@ struct iommu_hwpt_alloc {
>   enum iommu_hw_info_type {
>   	IOMMU_HW_INFO_TYPE_DEFAULT,
>   };
> +
> +/**
> + * struct iommu_hw_info - ioctl(IOMMU_DEVICE_GET_HW_INFO)
> + * @size: sizeof(struct iommu_hw_info)
> + * @flags: Must be 0
> + * @dev_id: The device being attached to the iommufd
> + * @data_len: Input the length of the user buffer in bytes. Output the
> + *            length of data filled to the user buffer.
> + * @data_ptr: Pointer to the type specific structure
> + * @out_data_type: Output the iommu hardware info type, it is one of
> + *                 enum iommu_hw_info_type.
> + * @__reserved: Must be 0
> + *
> + * Query the hardware iommu information for given device which has been
> + * bound to iommufd. @data_len is the size of the buffer which captures
> + * iommu type specific data and the data will be filled. Trailing bytes
> + * are zeroed if the user buffer is larger than the data kernel has.
> + *
> + * The type specific data would be used to sync capability between the
> + * vIOMMU and the hardware IOMMU. e.g. nested translation requires to
> + * check the hardware IOMMU capability, since a stage-1 translation table
> + * is owned by user but used by hardware IOMMU.
> + *
> + * The @out_data_type will be filled if the ioctl succeeds. It would
> + * be used to decode the data filled in the buffer pointed by @data_ptr.
> + *
> + * This is only available for the physical devices bound to iommufd as
> + * only physical devices can have hardware IOMMU.
> + */
> +struct iommu_hw_info {
> +	__u32 size;
> +	__u32 flags;
> +	__u32 dev_id;
> +	__u32 data_len;
> +	__aligned_u64 data_ptr;
> +	__u32 out_data_type;
> +	__u32 __reserved;
> +};
> +#define IOMMU_DEVICE_GET_HW_INFO _IO(IOMMUFD_TYPE, IOMMUFD_CMD_DEVICE_GET_HW_INFO)
>   #endif

Other look good to me.

Best regards,
baolu
  

On Thu, Mar 09, 2023 at 09:50:18PM +0800, Baolu Lu wrote:

> > +	if (cmd->flags || cmd->__reserved || !cmd->data_len)
> > +		return -EOPNOTSUPP;
> > +
> > +	idev = iommufd_get_device(ucmd, cmd->dev_id);
> > +	if (IS_ERR(idev))
> > +		return PTR_ERR(idev);
> > +
> > +	ops = dev_iommu_ops(idev->dev);
> > +	if (!ops || !ops->hw_info) {
> 
> dev_iommu_ops() will never return a NULL.
> 
> Need below check
> 
> 	dev->iommu && dev->iommu->iommu_dev
> 
> before dev_iommu_ops(). Perhaps something like below?
> 
> 	if (!dev->iommu || !dev->iommu->iommu_dev)
> 		return -EINVAL;

At this point the device has become owned through the ownership API,
it absolutely has to have an iommu and an ops. No need to check
anything.

Jason
  

> From: Jason Gunthorpe <jgg@nvidia.com>
> Sent: Friday, March 10, 2023 1:21 AM
> 
> On Thu, Mar 09, 2023 at 09:50:18PM +0800, Baolu Lu wrote:
> 
> > > +	if (cmd->flags || cmd->__reserved || !cmd->data_len)
> > > +		return -EOPNOTSUPP;
> > > +
> > > +	idev = iommufd_get_device(ucmd, cmd->dev_id);
> > > +	if (IS_ERR(idev))
> > > +		return PTR_ERR(idev);
> > > +
> > > +	ops = dev_iommu_ops(idev->dev);
> > > +	if (!ops || !ops->hw_info) {
> >
> > dev_iommu_ops() will never return a NULL.
> >
> > Need below check
> >
> > 	dev->iommu && dev->iommu->iommu_dev
> >
> > before dev_iommu_ops(). Perhaps something like below?
> >
> > 	if (!dev->iommu || !dev->iommu->iommu_dev)
> > 		return -EINVAL;
> 
> At this point the device has become owned through the ownership API,
> it absolutely has to have an iommu and an ops. No need to check
> anything.

ok. so just needs to check hw_info callback.
  

> From: Liu, Yi L <yi.l.liu@intel.com>
> Sent: Thursday, March 9, 2023 3:54 PM
> +
> +int iommufd_device_get_hw_info(struct iommufd_ucmd *ucmd)
> +{
> +	struct iommu_hw_info *cmd = ucmd->cmd;
> +	struct iommufd_device *idev;
> +	const struct iommu_ops *ops;
> +	void *data;
> +	unsigned int length, data_len;
> +	int rc;
> +
> +	if (cmd->flags || cmd->__reserved || !cmd->data_len)
> +		return -EOPNOTSUPP;
> +
> +	idev = iommufd_get_device(ucmd, cmd->dev_id);
> +	if (IS_ERR(idev))
> +		return PTR_ERR(idev);
> +
> +	ops = dev_iommu_ops(idev->dev);
> +	if (!ops || !ops->hw_info) {
> +		rc = -EOPNOTSUPP;
> +		goto out_put;
> +	}
> +
> +	/* driver has hw_info callback should have a unique driver_type */
> +	if (WARN_ON(ops->driver_type ==
> IOMMU_HW_INFO_TYPE_DEFAULT)) {
> +		rc = -EOPNOTSUPP;
> +		goto out_put;
> +	}

ok, here is where the check is done.

> +
> +	data = ops->hw_info(idev->dev, &data_len);

if we directly return type in @hw_info, this becomes:

	data = ops->hw_info(idev->dev, &data_len, &driver_type);

> +/**
> + * struct iommu_hw_info - ioctl(IOMMU_DEVICE_GET_HW_INFO)
> + * @size: sizeof(struct iommu_hw_info)
> + * @flags: Must be 0
> + * @dev_id: The device being attached to the iommufd

"The device bound to the iommufd"

> + * @data_len: Input the length of the user buffer in bytes. Output the
> + *            length of data filled to the user buffer.

s/to/in/

> + * @data_ptr: Pointer to the type specific structure

"Pointer to the user buffer"

> + * @out_data_type: Output the iommu hardware info type, it is one of
> + *                 enum iommu_hw_info_type.

s/it is one of/as defined by/

> + * @__reserved: Must be 0
> + *
> + * Query the hardware iommu information for given device which has been
> + * bound to iommufd. @data_len is the size of the buffer which captures
> + * iommu type specific data and the data will be filled. Trailing bytes
> + * are zeroed if the user buffer is larger than the data kernel has.
> + *
> + * The type specific data would be used to sync capability between the
> + * vIOMMU and the hardware IOMMU. e.g. nested translation requires to

s/vIOMMU/virtual IOMMU/

> + * check the hardware IOMMU capability, since a stage-1 translation table
> + * is owned by user but used by hardware IOMMU.

"check ... capability so guest stage-1 page table uses a format compatible
to the hardware IOMMU"

> + *
> + * The @out_data_type will be filled if the ioctl succeeds. It would
> + * be used to decode the data filled in the buffer pointed by @data_ptr.
> + *
> + * This is only available for the physical devices bound to iommufd as
> + * only physical devices can have hardware IOMMU.

not required. User doesn't know whether it's physical or emulated
device.