Re: [PATCH v3 2/9] vfio/fsl-mc: Scan DPRC objects on vfio-fsl-mc driver bind

[Alex Williamson]

On Mon,  6 Jul 2020 18:41:46 +0300
Diana Craciun <diana.craciun@xxxxxxxxxxx> wrote:

> The DPRC (Data Path Resource Container) device is a bus device and has
> child devices attached to it. When the vfio-fsl-mc driver is probed
> the DPRC is scanned and the child devices discovered and initialized.
> 
> Signed-off-by: Bharat Bhushan <Bharat.Bhushan@xxxxxxx>
> Signed-off-by: Diana Craciun <diana.craciun@xxxxxxxxxxx>
> ---
>  drivers/vfio/fsl-mc/vfio_fsl_mc.c         | 106 ++++++++++++++++++++++
>  drivers/vfio/fsl-mc/vfio_fsl_mc_private.h |   1 +
>  2 files changed, 107 insertions(+)
> 
> diff --git a/drivers/vfio/fsl-mc/vfio_fsl_mc.c b/drivers/vfio/fsl-mc/vfio_fsl_mc.c
> index 8b53c2a25b32..ad8d06cceb71 100644
> --- a/drivers/vfio/fsl-mc/vfio_fsl_mc.c
> +++ b/drivers/vfio/fsl-mc/vfio_fsl_mc.c
> @@ -15,6 +15,8 @@
>  
>  #include "vfio_fsl_mc_private.h"
>  
> +static struct fsl_mc_driver vfio_fsl_mc_driver;
> +
>  static int vfio_fsl_mc_open(void *device_data)
>  {
>  	if (!try_module_get(THIS_MODULE))
> @@ -84,6 +86,69 @@ static const struct vfio_device_ops vfio_fsl_mc_ops = {
>  	.mmap		= vfio_fsl_mc_mmap,
>  };
>  
> +static int vfio_fsl_mc_bus_notifier(struct notifier_block *nb,
> +				    unsigned long action, void *data)
> +{
> +	struct vfio_fsl_mc_device *vdev = container_of(nb,
> +					struct vfio_fsl_mc_device, nb);
> +	struct device *dev = data;
> +	struct fsl_mc_device *mc_dev = to_fsl_mc_device(dev);
> +	struct fsl_mc_device *mc_cont = to_fsl_mc_device(mc_dev->dev.parent);
> +
> +	if (action == BUS_NOTIFY_ADD_DEVICE &&
> +	    vdev->mc_dev == mc_cont) {
> +		mc_dev->driver_override = kasprintf(GFP_KERNEL, "%s",
> +						    vfio_fsl_mc_ops.name);

I notice the vfio-pci code that this is modeled from also doesn't check
this allocation for NULL.  Maybe both should print a dev_warn on the
ultra slim chance it would fail.

> +		dev_info(dev, "Setting driver override for device in dprc %s\n",
> +			 dev_name(&mc_cont->dev));
> +	} else if (action == BUS_NOTIFY_BOUND_DRIVER &&
> +		vdev->mc_dev == mc_cont) {
> +		struct fsl_mc_driver *mc_drv = to_fsl_mc_driver(dev->driver);
> +
> +		if (mc_drv && mc_drv != &vfio_fsl_mc_driver)
> +			dev_warn(dev, "Object %s bound to driver %s while DPRC bound to vfio-fsl-mc\n",
> +				 dev_name(dev), mc_drv->driver.name);
> +		}

Nit, } is over-indented, should be aligned to the previous 'else if'.

> +
> +	return 0;
> +}
> +
> +static int vfio_fsl_mc_init_device(struct vfio_fsl_mc_device *vdev)
> +{
> +	struct fsl_mc_device *mc_dev = vdev->mc_dev;
> +	int ret;
> +
> +	/* Non-dprc devices share mc_io from parent */
> +	if (!is_fsl_mc_bus_dprc(mc_dev)) {
> +		struct fsl_mc_device *mc_cont = to_fsl_mc_device(mc_dev->dev.parent);
> +
> +		mc_dev->mc_io = mc_cont->mc_io;
> +		return 0;
> +	}
> +
> +	vdev->nb.notifier_call = vfio_fsl_mc_bus_notifier;
> +	ret = bus_register_notifier(&fsl_mc_bus_type, &vdev->nb);
> +	if (ret)
> +		return ret;
> +
> +	/* open DPRC, allocate a MC portal */
> +	ret = dprc_setup(mc_dev);
> +	if (ret < 0) {
> +		dev_err(&mc_dev->dev, "Failed to setup DPRC (error = %d)\n", ret);
> +		bus_unregister_notifier(&fsl_mc_bus_type, &vdev->nb);
> +		return ret;
> +	}
> +
> +	ret = dprc_scan_container(mc_dev, false);
> +	if (ret < 0) {
> +		dev_err(&mc_dev->dev, "Container scanning failed: %d\n", ret);
> +		bus_unregister_notifier(&fsl_mc_bus_type, &vdev->nb);
> +		dprc_cleanup(mc_dev);

All else being equal, should these be reversed to mirror the setup?

> +	}
> +
> +	return ret;
> +}
> +
>  static int vfio_fsl_mc_probe(struct fsl_mc_device *mc_dev)
>  {
>  	struct iommu_group *group;
> @@ -112,9 +177,42 @@ static int vfio_fsl_mc_probe(struct fsl_mc_device *mc_dev)
>  		return ret;
>  	}
>  
> +	ret = vfio_fsl_mc_init_device(vdev);
> +	if (ret < 0) {
> +		vfio_iommu_group_put(group, dev);
> +		return ret;
> +	}
> +
>  	return ret;
>  }
>  
> +static int vfio_fsl_mc_device_remove(struct device *dev, void *data)
> +{
> +	struct fsl_mc_device *mc_dev;
> +
> +	WARN_ON(!dev);
> +	mc_dev = to_fsl_mc_device(dev);
> +	if (WARN_ON(!mc_dev))
> +		return -ENODEV;
> +
> +	kfree(mc_dev->driver_override);
> +	mc_dev->driver_override = NULL;

This is out of scope, all other buses that support a driver_override
free this is the bus driver code.  Why isn't it sufficient that it's
done in fsl_mc_device_remove()?

> +
> +	/*
> +	 * The device-specific remove callback will get invoked by device_del()
> +	 */
> +	device_del(&mc_dev->dev);
> +	put_device(&mc_dev->dev);

In fact, why are we doing any of this?  I think these devices were
created via dprc_scan_container(), so shouldn't there be a dprc
callback to remove them?  What happens if one of them did get bound to
another driver, haven't we just deleted the device out from under them?
In vfio-pci for instance, we call pci_disable_sriov() to remove any
vfs.  Thanks,

Alex

> +
> +	return 0;
> +}
> +
> +static void vfio_fsl_mc_cleanup_dprc(struct fsl_mc_device *mc_dev)
> +{
> +	device_for_each_child(&mc_dev->dev, NULL, vfio_fsl_mc_device_remove);
> +	dprc_cleanup(mc_dev);
> +}
> +
>  static int vfio_fsl_mc_remove(struct fsl_mc_device *mc_dev)
>  {
>  	struct vfio_fsl_mc_device *vdev;
> @@ -124,6 +222,14 @@ static int vfio_fsl_mc_remove(struct fsl_mc_device *mc_dev)
>  	if (!vdev)
>  		return -EINVAL;
>  
> +	if (vdev->nb.notifier_call)
> +		bus_unregister_notifier(&fsl_mc_bus_type, &vdev->nb);
> +
> +	if (is_fsl_mc_bus_dprc(mc_dev))
> +		vfio_fsl_mc_cleanup_dprc(vdev->mc_dev);
> +
> +	mc_dev->mc_io = NULL;
> +
>  	vfio_iommu_group_put(mc_dev->dev.iommu_group, dev);
>  
>  	return 0;
> diff --git a/drivers/vfio/fsl-mc/vfio_fsl_mc_private.h b/drivers/vfio/fsl-mc/vfio_fsl_mc_private.h
> index e79cc116f6b8..37d61eaa58c8 100644
> --- a/drivers/vfio/fsl-mc/vfio_fsl_mc_private.h
> +++ b/drivers/vfio/fsl-mc/vfio_fsl_mc_private.h
> @@ -9,6 +9,7 @@
>  
>  struct vfio_fsl_mc_device {
>  	struct fsl_mc_device		*mc_dev;
> +	struct notifier_block        nb;
>  };
>  
>  #endif /* VFIO_FSL_MC_PRIVATE_H */