linux-renesas-soc.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Liviu Dudau <Liviu.Dudau@arm.com>
To: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Cc: dri-devel@lists.freedesktop.org,
	linux-renesas-soc@vger.kernel.org,
	Kieran Bingham <kieran.bingham@ideasonboard.com>,
	Daniel Vetter <daniel.vetter@intel.com>
Subject: Re: [PATCH/RFC 2/2] drm: rcar-du: Allow importing non-contiguous dma-buf with VSP
Date: Mon, 13 Nov 2017 11:53:14 +0000	[thread overview]
Message-ID: <20171113115314.GB26389@e110455-lin.cambridge.arm.com> (raw)
In-Reply-To: <20171113103228.23312-3-laurent.pinchart+renesas@ideasonboard.com>

On Mon, Nov 13, 2017 at 12:32:28PM +0200, Laurent Pinchart wrote:
> When the DU sources its frames from a VSP, it performs no memory access
> and thus has no requirements on imported dma-buf memory types. In
> particular the DU could import a physically non-contiguous buffer that
> would later be mapped contiguously through the VSP IOMMU.
> 
> This use case isn't supported at the moment as the GEM CMA helpers will
> reject any non-contiguous buffer, and the DU isn't connected to an IOMMU
> that can make the buffer contiguous for DMA. Fix this by implementing a
> custom .gem_prime_import_sg_table() operation that accepts all imported
> dma-buf regardless of the number of scatterlist entries.

This patch raises the question of why use CMA at all if you can accept
any kind of buffers.

> 
> Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
> ---
>  drivers/gpu/drm/rcar-du/rcar_du_drv.c |  2 +-
>  drivers/gpu/drm/rcar-du/rcar_du_kms.c | 39 +++++++++++++++++++++++++++++++++++
>  drivers/gpu/drm/rcar-du/rcar_du_kms.h |  7 +++++++
>  3 files changed, 47 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/rcar-du/rcar_du_drv.c b/drivers/gpu/drm/rcar-du/rcar_du_drv.c
> index 48c166f925a3..d999231f98c7 100644
> --- a/drivers/gpu/drm/rcar-du/rcar_du_drv.c
> +++ b/drivers/gpu/drm/rcar-du/rcar_du_drv.c
> @@ -289,7 +289,7 @@ static struct drm_driver rcar_du_driver = {
>  	.gem_prime_import	= drm_gem_prime_import,
>  	.gem_prime_export	= drm_gem_prime_export,
>  	.gem_prime_get_sg_table	= drm_gem_cma_prime_get_sg_table,
> -	.gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table,
> +	.gem_prime_import_sg_table = rcar_du_gem_prime_import_sg_table,
>  	.gem_prime_vmap		= drm_gem_cma_prime_vmap,
>  	.gem_prime_vunmap	= drm_gem_cma_prime_vunmap,
>  	.gem_prime_mmap		= drm_gem_cma_prime_mmap,
> diff --git a/drivers/gpu/drm/rcar-du/rcar_du_kms.c b/drivers/gpu/drm/rcar-du/rcar_du_kms.c
> index 566d1a948c8f..2dd0c2ba047d 100644
> --- a/drivers/gpu/drm/rcar-du/rcar_du_kms.c
> +++ b/drivers/gpu/drm/rcar-du/rcar_du_kms.c
> @@ -20,6 +20,7 @@
>  #include <drm/drm_gem_cma_helper.h>
>  #include <drm/drm_gem_framebuffer_helper.h>
>  
> +#include <linux/dma-buf.h>
>  #include <linux/of_graph.h>
>  #include <linux/wait.h>
>  
> @@ -148,6 +149,44 @@ const struct rcar_du_format_info *rcar_du_format_info(u32 fourcc)
>   * Frame buffer
>   */
>  
> +struct drm_gem_object *rcar_du_gem_prime_import_sg_table(struct drm_device *dev,
> +				struct dma_buf_attachment *attach,
> +				struct sg_table *sgt)
> +{
> +	struct rcar_du_device *rcdu = dev->dev_private;
> +	struct drm_gem_cma_object *cma_obj;
> +	struct drm_gem_object *gem_obj;
> +	int ret;
> +
> +	if (!rcar_du_has(rcdu, RCAR_DU_FEATURE_VSP1_SOURCE))
> +		return drm_gem_cma_prime_import_sg_table(dev, attach, sgt);
> +
> +	/* Create a CMA GEM buffer. */
> +	cma_obj = kzalloc(sizeof(*cma_obj), GFP_KERNEL);
> +	if (!cma_obj)
> +		return ERR_PTR(-ENOMEM);
> +	gem_obj = &cma_obj->base;
> +
> +	ret = drm_gem_object_init(dev, gem_obj, attach->dmabuf->size);
> +	if (ret)
> +		goto error;
> +
> +	ret = drm_gem_create_mmap_offset(gem_obj);
> +	if (ret) {
> +		drm_gem_object_release(gem_obj);
> +		goto error;
> +	}
> +
> +	cma_obj->paddr = 0;

This is going to break drm_gem_cma_describe() if you are using it plus
the rcar_du_plane_setup_scanout() unless I'm missing something besides
familiarity with the RCAR driver code :)

This function looks very similar to what I tried to do for mali-dp to
allow the import of contiguous DMA buffers that have more than 1 sgt
entries. In the end I gave up as I kept finding issues and went for the
drm_gem_cma_prime_import_sg_table() changes. Maybe you need to do a
similar change in the function to bypass some requirements if the driver
signals that it can accept relaxed requirements?

Best regards,
Liviu

> +	cma_obj->sgt = sgt;
> +
> +	return gem_obj;
> +
> +error:
> +	kfree(cma_obj);
> +	return ERR_PTR(ret);
> +}
> +
>  int rcar_du_dumb_create(struct drm_file *file, struct drm_device *dev,
>  			struct drm_mode_create_dumb *args)
>  {
> diff --git a/drivers/gpu/drm/rcar-du/rcar_du_kms.h b/drivers/gpu/drm/rcar-du/rcar_du_kms.h
> index 07951d5fe38b..10b2bb0f0df9 100644
> --- a/drivers/gpu/drm/rcar-du/rcar_du_kms.h
> +++ b/drivers/gpu/drm/rcar-du/rcar_du_kms.h
> @@ -16,10 +16,13 @@
>  
>  #include <linux/types.h>
>  
> +struct dma_buf_attachment;
>  struct drm_file;
>  struct drm_device;
> +struct drm_gem_object;
>  struct drm_mode_create_dumb;
>  struct rcar_du_device;
> +struct sg_table;
>  
>  struct rcar_du_format_info {
>  	u32 fourcc;
> @@ -36,4 +39,8 @@ int rcar_du_modeset_init(struct rcar_du_device *rcdu);
>  int rcar_du_dumb_create(struct drm_file *file, struct drm_device *dev,
>  			struct drm_mode_create_dumb *args);
>  
> +struct drm_gem_object *rcar_du_gem_prime_import_sg_table(struct drm_device *dev,
> +				struct dma_buf_attachment *attach,
> +				struct sg_table *sgt);
> +
>  #endif /* __RCAR_DU_KMS_H__ */
> -- 
> Regards,
> 
> Laurent Pinchart
> 

-- 
====================
| I would like to |
| fix the world,  |
| but they're not |
| giving me the   |
 \ source code!  /
  ---------------
    ¯\_(ツ)_/¯

  reply	other threads:[~2017-11-13 11:53 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-11-13 10:32 [PATCH/RFC 0/2] R-Car DU: Support importing non-contiguous dma-buf with VSP Laurent Pinchart
2017-11-13 10:32 ` [PATCH/RFC 1/2] drm: rcar-du: Set the DMA coherent mask for the DU device Laurent Pinchart
2017-11-13 11:44   ` Liviu Dudau
2017-11-13 10:32 ` [PATCH/RFC 2/2] drm: rcar-du: Allow importing non-contiguous dma-buf with VSP Laurent Pinchart
2017-11-13 11:53   ` Liviu Dudau [this message]
2017-11-14  3:34     ` Laurent Pinchart
2017-11-14  9:27       ` Liviu Dudau
2017-12-13 22:46   ` [PATCH/RFC v1.1] " Laurent Pinchart
2017-11-13 13:48 ` [PATCH/RFC 0/2] R-Car DU: Support " Kieran Bingham
2018-12-18  7:40   ` Laurent Pinchart

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20171113115314.GB26389@e110455-lin.cambridge.arm.com \
    --to=liviu.dudau@arm.com \
    --cc=daniel.vetter@intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=kieran.bingham@ideasonboard.com \
    --cc=laurent.pinchart+renesas@ideasonboard.com \
    --cc=linux-renesas-soc@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).