[PATCH] media: atomisp: ssh_css: Fix a null-pointer dereference in load_video_binaries

[PATCH] media: atomisp: ssh_css: Fix a null-pointer dereference in load_video_binaries

[Zhipeng Lu]

The allocation failure of mycs->yuv_scaler_binary in load_video_binaries
is followed with a dereference of mycs->yuv_scaler_binary after the
following call chain:

sh_css_pipe_load_binaries
  |-> load_video_binaries (mycs->yuv_scaler_binary == NULL)
  |
  |-> sh_css_pipe_unload_binaries
        |-> unload_video_binaries

In unload_video_binaries, it calls to ia_css_binary_unload with argument
&pipe->pipe_settings.video.yuv_scaler_binary[i], which refers to the
same memory slot as mycs->yuv_scaler_binary. Thus, a null-pointer
dereference is triggered.

Fixes: ad85094b293e ("Revert "media: staging: atomisp: Remove driver"")
Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn>
---
 drivers/staging/media/atomisp/pci/sh_css.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/staging/media/atomisp/pci/sh_css.c b/drivers/staging/media/atomisp/pci/sh_css.c
index f35c90809414..eb43f4e99d02 100644
--- a/drivers/staging/media/atomisp/pci/sh_css.c
+++ b/drivers/staging/media/atomisp/pci/sh_css.c
@@ -4936,9 +4936,10 @@ unload_video_binaries(struct ia_css_pipe *pipe)
 	ia_css_binary_unload(&pipe->pipe_settings.video.video_binary);
 	ia_css_binary_unload(&pipe->pipe_settings.video.vf_pp_binary);
 
-	for (i = 0; i < pipe->pipe_settings.video.num_yuv_scaler; i++)
-		ia_css_binary_unload(&pipe->pipe_settings.video.yuv_scaler_binary[i]);
-
+	if (pipe->pipe_settings.video.yuv_scaler_binary)
+		for (i = 0; i < pipe->pipe_settings.video.num_yuv_scaler; i++)
+			ia_css_binary_unload(&pipe->pipe_settings.video.yuv_scaler_binary[i]);
+		
 	kfree(pipe->pipe_settings.video.is_output_stage);
 	pipe->pipe_settings.video.is_output_stage = NULL;
 	kfree(pipe->pipe_settings.video.yuv_scaler_binary);
-- 
2.34.1

Re: [PATCH] media: atomisp: ssh_css: Fix a null-pointer dereference in load_video_binaries

[Hans de Goede]

Hi Zhipeng Lu,

On 1/12/24 09:34, Zhipeng Lu wrote:
> The allocation failure of mycs->yuv_scaler_binary in load_video_binaries
> is followed with a dereference of mycs->yuv_scaler_binary after the
> following call chain:
> 
> sh_css_pipe_load_binaries
>   |-> load_video_binaries (mycs->yuv_scaler_binary == NULL)
>   |
>   |-> sh_css_pipe_unload_binaries
>         |-> unload_video_binaries
> 
> In unload_video_binaries, it calls to ia_css_binary_unload with argument
> &pipe->pipe_settings.video.yuv_scaler_binary[i], which refers to the
> same memory slot as mycs->yuv_scaler_binary. Thus, a null-pointer
> dereference is triggered.
> 
> Fixes: ad85094b293e ("Revert "media: staging: atomisp: Remove driver"")
> Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn>

Thank you for your patch. I believe it would be better to fix this
like this:

diff --git a/drivers/staging/media/atomisp/pci/sh_css.c b/drivers/staging/media/atomisp/pci/sh_css.c
index 1d1fbda75da1..d566c5417448 100644
--- a/drivers/staging/media/atomisp/pci/sh_css.c
+++ b/drivers/staging/media/atomisp/pci/sh_css.c
@@ -4690,6 +4690,7 @@ static int load_video_binaries(struct ia_css_pipe *pipe)
 						  sizeof(struct ia_css_binary),
 						  GFP_KERNEL);
 		if (!mycs->yuv_scaler_binary) {
+			mycs->num_yuv_scaler = 0;
 			err = -ENOMEM;
 			return err;
 		}

Can you please submit a new version using this approach ?

Regards,

Hans



> ---
>  drivers/staging/media/atomisp/pci/sh_css.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/staging/media/atomisp/pci/sh_css.c b/drivers/staging/media/atomisp/pci/sh_css.c
> index f35c90809414..eb43f4e99d02 100644
> --- a/drivers/staging/media/atomisp/pci/sh_css.c
> +++ b/drivers/staging/media/atomisp/pci/sh_css.c
> @@ -4936,9 +4936,10 @@ unload_video_binaries(struct ia_css_pipe *pipe)
>  	ia_css_binary_unload(&pipe->pipe_settings.video.video_binary);
>  	ia_css_binary_unload(&pipe->pipe_settings.video.vf_pp_binary);
>  
> -	for (i = 0; i < pipe->pipe_settings.video.num_yuv_scaler; i++)
> -		ia_css_binary_unload(&pipe->pipe_settings.video.yuv_scaler_binary[i]);
> -
> +	if (pipe->pipe_settings.video.yuv_scaler_binary)
> +		for (i = 0; i < pipe->pipe_settings.video.num_yuv_scaler; i++)
> +			ia_css_binary_unload(&pipe->pipe_settings.video.yuv_scaler_binary[i]);
> +		
>  	kfree(pipe->pipe_settings.video.is_output_stage);
>  	pipe->pipe_settings.video.is_output_stage = NULL;
>  	kfree(pipe->pipe_settings.video.yuv_scaler_binary);

Re: [PATCH] media: atomisp: ssh_css: Fix a null-pointer dereference in load_video_binaries

[Andy Shevchenko]

On Fri, Jan 12, 2024 at 10:49 AM Hans de Goede <hdegoede@redhat.com> wrote:
> On 1/12/24 09:34, Zhipeng Lu wrote:


> > Fixes: ad85094b293e ("Revert "media: staging: atomisp: Remove driver"")

> Can you please submit a new version using this approach ?

Besides that, are you sure the Fixes refers to the correct commit?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] media: atomisp: ssh_css: Fix a null-pointer dereference in load_video_binaries

[Andy Shevchenko]

On Wed, Jan 17, 2024 at 8:34 AM <alexious@zju.edu.cn> wrote:
> > On Fri, Jan 12, 2024 at 10:49 AM Hans de Goede <hdegoede@redhat.com> wrote:
> > > On 1/12/24 09:34, Zhipeng Lu wrote:

> > > > Fixes: ad85094b293e ("Revert "media: staging: atomisp: Remove driver"")
> > Besides that, are you sure the Fixes refers to the correct commit?
>
> Well, I think I referred to the correct commit, which introduce the whole module and leave this bug.
>
> If I did miss something please let me know.

Yes, the driver was before that commit in the kernel. Was it without
the bug? No, because you are referring to a clear revert. So, find the
real commit that had brought that into the kernel.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] media: atomisp: ssh_css: Fix a null-pointer dereference in load_video_binaries

[alexious]

> On Wed, Jan 17, 2024 at 8:34 AM <alexious@zju.edu.cn> wrote:
> > > On Fri, Jan 12, 2024 at 10:49 AM Hans de Goede <hdegoede@redhat.com> wrote:
> > > > On 1/12/24 09:34, Zhipeng Lu wrote:
> 
> > > > > Fixes: ad85094b293e ("Revert "media: staging: atomisp: Remove driver"")
> > > Besides that, are you sure the Fixes refers to the correct commit?
> >
> > Well, I think I referred to the correct commit, which introduce the whole module and leave this bug.
> >
> > If I did miss something please let me know.
> 
> Yes, the driver was before that commit in the kernel. Was it without
> the bug? No, because you are referring to a clear revert. So, find the
> real commit that had brought that into the kernel.

You are correct, I just did some git blame on the latest version but 
forgot about the commit blamed was a revert commit.
A v2 version of this patch will be sent later to fix this issue.
Thank you for pointing out my mistake.

Regards,
Zhipeng

> 
> -- 
> With Best Regards,
> Andy Shevchenko