Re: [PATCH v3] media: rga: fix NULL pointer dereferences, use-after-free, memory leak

[Steven Price <steven.price@arm.com>]

On 14/03/2019 05:03, Kangjie Lu wrote:
> 1. dma_alloc_attrs, __get_free_pages can fail and return NULL.
> Further uses of their return values lead to NULL pointer
> dereferences
> 
> 2. In the error-handling path, video_unregister_device uses
> "rga->vfd" which has been freed by video_device_release
> 
> 3. The error handling for v4l2_m2m_init and video_register_device
> has memory-leak issues
> 
> The patch fixes the above issues.
> 
> Signed-off-by: Kangjie Lu <kjlu@umn.edu>

Reviewed-by: Steven Price <steven.price@arm.com>

Yes, that looks like it solves the issues with the clean up - thanks for
reworking it!

> ---
>  drivers/media/platform/rockchip/rga/rga.c | 26 ++++++++++++++++++++---
>  1 file changed, 23 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/media/platform/rockchip/rga/rga.c b/drivers/media/platform/rockchip/rga/rga.c
> index 5c653287185f..468365ceb99d 100644
> --- a/drivers/media/platform/rockchip/rga/rga.c
> +++ b/drivers/media/platform/rockchip/rga/rga.c
> @@ -889,11 +889,24 @@ static int rga_probe(struct platform_device *pdev)
>  	rga->cmdbuf_virt = dma_alloc_attrs(rga->dev, RGA_CMDBUF_SIZE,
>  					   &rga->cmdbuf_phy, GFP_KERNEL,
>  					   DMA_ATTR_WRITE_COMBINE);
> +	if (!rga->cmdbuf_virt) {
> +		ret = -ENOMEM;
> +		goto unreg_video_dev;
> +	}
>  
>  	rga->src_mmu_pages =
>  		(unsigned int *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 3);
> +	if (!rga->src_mmu_pages) {
> +		ret = -ENOMEM;
> +		goto free_dma_attrs;
> +	}
> +
>  	rga->dst_mmu_pages =
>  		(unsigned int *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 3);
> +	if (!rga->dst_mmu_pages) {
> +		ret = -ENOMEM;
> +		goto free_dst_pages;
> +	}
>  
>  	def_frame.stride = (def_frame.width * def_frame.fmt->depth) >> 3;
>  	def_frame.size = def_frame.stride * def_frame.height;
> @@ -901,7 +914,7 @@ static int rga_probe(struct platform_device *pdev)
>  	ret = video_register_device(vfd, VFL_TYPE_GRABBER, -1);
>  	if (ret) {
>  		v4l2_err(&rga->v4l2_dev, "Failed to register video device\n");
> -		goto rel_vdev;
> +		goto free_pages;
>  	}
>  
>  	v4l2_info(&rga->v4l2_dev, "Registered %s as /dev/%s\n",
> @@ -909,10 +922,17 @@ static int rga_probe(struct platform_device *pdev)
>  
>  	return 0;
>  
> -rel_vdev:
> -	video_device_release(vfd);
> +free_pages:
> +	free_pages((unsigned long)rga->src_mmu_pages, 3);
> +free_dst_pages:
> +	free_pages((unsigned long)rga->dst_mmu_pages, 3);

Minor comment: free_pages accepts a NULL pointer (or more precisely 0,
since it takes an unsigned long), it just returns having done nothing.
So there isn't actually any harm in calling free_pages unconditionally
in the error path (since rga is initialised to all zeros). You can save
a couple of labels in the error path using that trick.

Thanks,

Steve

> +free_dma_attrs:
> +	dma_free_attrs(rga->dev, RGA_CMDBUF_SIZE, rga->cmdbuf_virt,
> +		       rga->cmdbuf_phy,
> +		       DMA_ATTR_WRITE_COMBINE);
>  unreg_video_dev:
>  	video_unregister_device(rga->vfd);
> +	video_device_release(vfd);
>  unreg_v4l2_dev:
>  	v4l2_device_unregister(&rga->v4l2_dev);
>  err_put_clk:
>