[PATCH v5] drm: Use USB controller's DMA mask when importing dmabufs

[Greg KH]

On Fri, Feb 26, 2021 at 10:26:47AM +0100, Thomas Zimmermann wrote:
> USB devices cannot perform DMA and hence have no dma_mask set in their
> device structure. Therefore importing dmabuf into a USB-based driver
> fails, which breaks joining and mirroring of display in X11.
> 
> For USB devices, pick the associated USB controller as attachment device.
> This allows the DRM import helpers to perform the DMA setup. If the DMA
> controller does not support DMA transfers, we're out of luck and cannot
> import. Our current USB-based DRM drivers don't use DMA, so the actual
> DMA device is not important.
> 
> Drivers should use DRM_GEM_SHMEM_DROVER_OPS_USB to initialize their
> instance of struct drm_driver.
> 
> Tested by joining/mirroring displays of udl and radeon un der Gnome/X11.
> 
> v5:
> 	* provide a helper for USB interfaces (Alan)
> 	* add FIXME item to documentation and TODO list (Daniel)
> v4:
> 	* implement workaround with USB helper functions (Greg)
> 	* use struct usb_device->bus->sysdev as DMA device (Takashi)
> v3:
> 	* drop gem_create_object
> 	* use DMA mask of USB controller, if any (Daniel, Christian, Noralf)
> v2:
> 	* move fix to importer side (Christian, Daniel)
> 	* update SHMEM and CMA helpers for new PRIME callbacks
> 
> Signed-off-by: Thomas Zimmermann <tzimmermann at suse.de>
> Fixes: 6eb0233ec2d0 ("usb: don't inherity DMA properties for USB devices")
> Tested-by: Pavel Machek <pavel at ucw.cz>
> Acked-by: Christian König <christian.koenig at amd.com>
> Acked-by: Daniel Vetter <daniel.vetter at ffwll.ch>
> Cc: Christoph Hellwig <hch at lst.de>
> Cc: Greg Kroah-Hartman <gregkh at linuxfoundation.org>
> Cc: <stable at vger.kernel.org> # v5.10+
> ---
>  Documentation/gpu/todo.rst         | 15 ++++++++++
>  drivers/gpu/drm/drm_prime.c        | 45 ++++++++++++++++++++++++++++++
>  drivers/gpu/drm/tiny/gm12u320.c    |  2 +-
>  drivers/gpu/drm/udl/udl_drv.c      |  2 +-
>  drivers/usb/core/usb.c             | 31 ++++++++++++++++++++
>  include/drm/drm_gem_shmem_helper.h | 16 +++++++++++
>  include/drm/drm_prime.h            |  5 ++++
>  include/linux/usb.h                | 24 ++++++++++++++++
>  8 files changed, 138 insertions(+), 2 deletions(-)
> 
> diff --git a/Documentation/gpu/todo.rst b/Documentation/gpu/todo.rst
> index f872d3d33218..c185e0a2951e 100644
> --- a/Documentation/gpu/todo.rst
> +++ b/Documentation/gpu/todo.rst
> @@ -617,6 +617,21 @@ Contact: Daniel Vetter
>  
>  Level: Intermediate
>  
> +Remove automatic page mapping from dma-buf importing
> +----------------------------------------------------
> +
> +When importing dma-bufs, the dma-buf and PRIME frameworks automatically map
> +imported pages into the importer's DMA area. This is a problem for USB devices,
> +which do not support DMA operations. By default, importing fails for USB
> +devices. USB-based drivers work around this problem by employing
> +drm_gem_prime_import_usb(). To fix the issue, automatic page mappings should
> +be removed from the buffer-sharing code.
> +
> +Contact: Thomas Zimmermann <tzimmermann at suse.de>, Daniel Vetter
> +
> +Level: Advanced
> +
> +
>  Better Testing
>  ==============
>  
> diff --git a/drivers/gpu/drm/drm_prime.c b/drivers/gpu/drm/drm_prime.c
> index 2a54f86856af..59013bb1cd4b 100644
> --- a/drivers/gpu/drm/drm_prime.c
> +++ b/drivers/gpu/drm/drm_prime.c
> @@ -29,6 +29,7 @@
>  #include <linux/export.h>
>  #include <linux/dma-buf.h>
>  #include <linux/rbtree.h>
> +#include <linux/usb.h>
>  
>  #include <drm/drm.h>
>  #include <drm/drm_drv.h>
> @@ -1055,3 +1056,47 @@ void drm_prime_gem_destroy(struct drm_gem_object *obj, struct sg_table *sg)
>  	dma_buf_put(dma_buf);
>  }
>  EXPORT_SYMBOL(drm_prime_gem_destroy);
> +
> +/**
> + * drm_gem_prime_import_usb - helper library implementation of the import callback for USB devices
> + * @dev: drm_device to import into
> + * @dma_buf: dma-buf object to import
> + *
> + * This is an implementation of drm_gem_prime_import() for USB-based devices.
> + * USB devices cannot perform DMA directly. This function selects the USB host
> + * controller as DMA device instead. Drivers can use this as their
> + * &drm_driver.gem_prime_import implementation.
> + *
> + * See also drm_gem_prime_import().
> + *
> + * FIXME: The dma-buf framework expects to map the exported pages into
> + *        the importer's DMA area. USB devices don't support DMA, and
> + *        importing would fail. Foir the time being, this function provides
> + *        a workaround by using the USB controller's DMA area. The real
> + *        solution is to remove page-mapping operations from the dma-buf
> + *        framework.
> + *
> + * Returns: A GEM object on success, or a pointer-encoder errno value otherwise.
> + */
> +#ifdef CONFIG_USB
> +struct drm_gem_object *drm_gem_prime_import_usb(struct drm_device *dev,
> +						struct dma_buf *dma_buf)
> +{
> +	struct device *dmadev;
> +	struct drm_gem_object *obj;
> +
> +	if (!dev_is_usb(dev->dev))
> +		return ERR_PTR(-ENODEV);

I have resisted the "dev_is_*()" type of function for USB for a long
time now, and I really don't want to add it now.

The driver core explicitly was not created with RTI (run type
identification), but over time it has been slowly added on a per-bus
basis for various reasons, some good and others not good.

In this function, why would a drm device that was NOT a usb device ever
call it?  Because of that, I don't think dev_is_usb() is needed at all,
just don't call this function unless it really is a USB device.

If you need help enforcing it, add a 'struct usb_interface *' to the
function parameters but please don't create this function.

Hm, looks like we have 2 extern definitions of usb_bus_type in different
.h files, I should go fix that up now to try to prevent this type of
thing in the first place...

thanks,

greg k-h