[Intel-gfx] [PATCH] dma-buf: add ref counting for module as exporter

[Greg KH]

On Thu, May 07, 2015 at 01:00:52PM +0530, Sumit Semwal wrote:
> Add reference counting on a kernel module that exports dma-buf and
> implements its operations. This prevents the module from being unloaded
> while DMABUF file is in use.
> 
> The original patch [1] was submitted by Tomasz, but he's since shifted
> jobs and a ping didn't elicit any response.
> 
>   [tomasz: Original author]
> Signed-off-by: Tomasz Stanislawski <t.stanislaws at samsung.com>
> Acked-by: Daniel Vetter <daniel at ffwll.ch>
>   [sumits: updated & rebased as per review comments]
> Signed-off-by: Sumit Semwal <sumit.semwal at linaro.org>
> 
> [1]: https://lkml.org/lkml/2012/8/8/163
> ---
>  drivers/dma-buf/dma-buf.c                      | 9 ++++++++-
>  drivers/gpu/drm/armada/armada_gem.c            | 1 +
>  drivers/gpu/drm/drm_prime.c                    | 1 +
>  drivers/gpu/drm/exynos/exynos_drm_dmabuf.c     | 1 +
>  drivers/gpu/drm/i915/i915_gem_dmabuf.c         | 1 +
>  drivers/gpu/drm/omapdrm/omap_gem_dmabuf.c      | 1 +
>  drivers/gpu/drm/tegra/gem.c                    | 1 +
>  drivers/gpu/drm/udl/udl_dmabuf.c               | 1 +
>  drivers/gpu/drm/vmwgfx/vmwgfx_prime.c          | 1 +
>  drivers/media/v4l2-core/videobuf2-dma-contig.c | 1 +
>  drivers/media/v4l2-core/videobuf2-dma-sg.c     | 1 +
>  drivers/media/v4l2-core/videobuf2-vmalloc.c    | 1 +
>  drivers/staging/android/ion/ion.c              | 1 +
>  include/linux/dma-buf.h                        | 2 ++
>  14 files changed, 22 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c
> index c5a9138a6a8d..9583eac0238f 100644
> --- a/drivers/dma-buf/dma-buf.c
> +++ b/drivers/dma-buf/dma-buf.c
> @@ -29,6 +29,7 @@
>  #include <linux/anon_inodes.h>
>  #include <linux/export.h>
>  #include <linux/debugfs.h>
> +#include <linux/module.h>
>  #include <linux/seq_file.h>
>  #include <linux/poll.h>
>  #include <linux/reservation.h>
> @@ -64,6 +65,7 @@ static int dma_buf_release(struct inode *inode, struct file *file)
>  	BUG_ON(dmabuf->cb_shared.active || dmabuf->cb_excl.active);
>  
>  	dmabuf->ops->release(dmabuf);
> +	module_put(dmabuf->ops->owner);
>  
>  	mutex_lock(&db_list.lock);
>  	list_del(&dmabuf->list_node);
> @@ -302,9 +304,14 @@ struct dma_buf *dma_buf_export(const struct dma_buf_export_info *exp_info)
>  		return ERR_PTR(-EINVAL);
>  	}
>  
> +	if (!try_module_get(exp_info->ops->owner))
> +		return ERR_PTR(-ENOENT);
> +
>  	dmabuf = kzalloc(alloc_size, GFP_KERNEL);
> -	if (dmabuf == NULL)
> +	if (!dmabuf) {
> +		module_put(exp_info->ops->owner);
>  		return ERR_PTR(-ENOMEM);
> +	}
>  
>  	dmabuf->priv = exp_info->priv;
>  	dmabuf->ops = exp_info->ops;
> diff --git a/drivers/gpu/drm/armada/armada_gem.c b/drivers/gpu/drm/armada/armada_gem.c
> index 580e10acaa3a..d2c5fc1269b7 100644
> --- a/drivers/gpu/drm/armada/armada_gem.c
> +++ b/drivers/gpu/drm/armada/armada_gem.c
> @@ -524,6 +524,7 @@ armada_gem_dmabuf_mmap(struct dma_buf *buf, struct vm_area_struct *vma)
>  }
>  
>  static const struct dma_buf_ops armada_gem_prime_dmabuf_ops = {
> +	.owner = THIS_MODULE,
>  	.map_dma_buf	= armada_gem_prime_map_dma_buf,
>  	.unmap_dma_buf	= armada_gem_prime_unmap_dma_buf,
>  	.release	= drm_gem_dmabuf_release,

The "easier" way to do this is change the registration function to add
the module owner automatically, which keeps you from having to modify
all of the individual drivers.  Look at how pci and usb do this for
their driver registration functions.  That should result in a much
smaller patch, that always works properly for everyone (there's no way
for driver to get it wrong.)

thanks,

greg k-h