[PATCH] video: fbdev: imxfb: use after free in imxfb_remove()

We free "info" then dereference it on the next line.  I've moved the
call to framebuffer_release() down a line to avoid this problem.

Fixes: b7d2d37276c1 ("video: imxfb: Remove unused fields from platform data structure")
Signed-off-by: Dan Carpenter <redacted>

diff --git a/drivers/video/fbdev/imxfb.c b/drivers/video/fbdev/imxfb.c
index c166e0725be5..1d0c3592b290 100644
--- a/drivers/video/fbdev/imxfb.c
+++ b/drivers/video/fbdev/imxfb.c

@@ -1080,10 +1080,9 @@ static int imxfb_remove(struct platform_device *pdev)
 
 	fb_dealloc_cmap(&info->cmap);
 	kfree(info->pseudo_palette);
-	framebuffer_release(info);
-
 	dma_free_wc(&pdev->dev, fbi->map_size, info->screen_base,
 		    fbi->map_dma);
+	framebuffer_release(info);
 
 	iounmap(fbi->regs);
 	release_mem_region(res->start, resource_size(res));

Hi,

On Friday, June 30, 2017 11:02:42 AM Dan Carpenter wrote:

Good catch.

While you are at it, please:

- move framebuffer_release() after iounmap() (memory occupied by
  fbi will also be freed by framebuffer_release() call, please see
  framebuffer_alloc() for details)

- make the ordering in imxfb_remove() match the ordering of error
  path in imxfb_probe() (to improve consistency and prevent similar
  ordering errors in future)

 	release_mem_region(res->start, resource_size(res));

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

On Tue, Jul 04, 2017 at 05:44:57PM +0200, Bartlomiej Zolnierkiewicz wrote:

Hi,

On Friday, June 30, 2017 11:02:42 AM Dan Carpenter wrote:

quoted

We free "info" then dereference it on the next line.  I've moved the
call to framebuffer_release() down a line to avoid this problem.

Fixes: b7d2d37276c1 ("video: imxfb: Remove unused fields from platform data structure")
Signed-off-by: Dan Carpenter <redacted>

diff --git a/drivers/video/fbdev/imxfb.c b/drivers/video/fbdev/imxfb.c
index c166e0725be5..1d0c3592b290 100644
--- a/drivers/video/fbdev/imxfb.c
+++ b/drivers/video/fbdev/imxfb.c

@@ -1080,10 +1080,9 @@ static int imxfb_remove(struct platform_device *pdev)
 
 	fb_dealloc_cmap(&info->cmap);
 	kfree(info->pseudo_palette);
-	framebuffer_release(info);
-
 	dma_free_wc(&pdev->dev, fbi->map_size, info->screen_base,
 		    fbi->map_dma);
+	framebuffer_release(info);
 
 	iounmap(fbi->regs);

Good catch.

While you are at it, please:

- move framebuffer_release() after iounmap() (memory occupied by
  fbi will also be freed by framebuffer_release() call, please see
  framebuffer_alloc() for details)

I'm still recovering from an illness so I might not be thinking
straight.  I don't see a bug here.  But I agree that it should mirror
the probe function so I'll re-write the patch.

regards,
dan carpenter

On Wednesday, July 05, 2017 03:02:47 PM Dan Carpenter wrote:

On Tue, Jul 04, 2017 at 05:44:57PM +0200, Bartlomiej Zolnierkiewicz wrote:

quoted

Hi,

On Friday, June 30, 2017 11:02:42 AM Dan Carpenter wrote:

quoted

We free "info" then dereference it on the next line.  I've moved the
call to framebuffer_release() down a line to avoid this problem.

Fixes: b7d2d37276c1 ("video: imxfb: Remove unused fields from platform data structure")
Signed-off-by: Dan Carpenter <redacted>

diff --git a/drivers/video/fbdev/imxfb.c b/drivers/video/fbdev/imxfb.c
index c166e0725be5..1d0c3592b290 100644
--- a/drivers/video/fbdev/imxfb.c
+++ b/drivers/video/fbdev/imxfb.c

@@ -1080,10 +1080,9 @@ static int imxfb_remove(struct platform_device *pdev)
 
 	fb_dealloc_cmap(&info->cmap);
 	kfree(info->pseudo_palette);
-	framebuffer_release(info);
-
 	dma_free_wc(&pdev->dev, fbi->map_size, info->screen_base,
 		    fbi->map_dma);
+	framebuffer_release(info);
 
 	iounmap(fbi->regs);

Good catch.

While you are at it, please:

- move framebuffer_release() after iounmap() (memory occupied by
  fbi will also be freed by framebuffer_release() call, please see
  framebuffer_alloc() for details)

I'm still recovering from an illness so I might not be thinking
straight.  I don't see a bug here.  But I agree that it should mirror

Please take a look at the imxfb_probe():

...
	struct imxfb_info *fbi;
...
	info = framebuffer_alloc(sizeof(struct imxfb_info), &pdev->dev);
	if (!info)
		return -ENOMEM;

	fbi = info->par;
...

and framebuffer_alloc():

struct fb_info *framebuffer_alloc(size_t size, struct device *dev)
...
	int fb_info_size = sizeof(struct fb_info);
...
	p = kzalloc(fb_info_size + size, GFP_KERNEL);

	if (!p)
		return NULL;

	info = (struct fb_info *) p;

	if (size)
		info->par = p + fb_info_size;
...

the probe function so I'll re-write the patch.

Great!

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics