[PATCH v2 0/2] drm/shmem-helper: Fix a couple of error path bugs

[PATCH v2 0/2] drm/shmem-helper: Fix a couple of error path bugs

[Rob Clark]

From: Rob Clark <robdclark@chromium.org>

A couple fixes for error paths that userspace could manage to trigger.

Rob Clark (2):
  drm/shmem-helper: Remove errant put in error path
  drm/shmem-helper: Avoid vm_open error paths

 drivers/gpu/drm/drm_gem_shmem_helper.c | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

-- 
2.38.1

[PATCH v2 1/2] drm/shmem-helper: Remove errant put in error path

[Rob Clark]

From: Rob Clark <robdclark@chromium.org>

drm_gem_shmem_mmap() doesn't own this reference, resulting in the GEM
object getting prematurely freed leading to a later use-after-free.

Link: https://syzkaller.appspot.com/bug?extid=c8ae65286134dd1b800d
Reported-by: syzbot+c8ae65286134dd1b800d@syzkaller.appspotmail.com
Fixes: 2194a63a818d ("drm: Add library for shmem backed GEM objects")
Cc: stable@vger.kernel.org
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/drm_gem_shmem_helper.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem_shmem_helper.c b/drivers/gpu/drm/drm_gem_shmem_helper.c
index 35138f8a375c..3b7b71391a4c 100644
--- a/drivers/gpu/drm/drm_gem_shmem_helper.c
+++ b/drivers/gpu/drm/drm_gem_shmem_helper.c
@@ -622,10 +622,8 @@ int drm_gem_shmem_mmap(struct drm_gem_shmem_object *shmem, struct vm_area_struct
 	}
 
 	ret = drm_gem_shmem_get_pages(shmem);
-	if (ret) {
-		drm_gem_vm_close(vma);
+	if (ret)
 		return ret;
-	}
 
 	vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
 	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
-- 
2.38.1

[PATCH v2 2/2] drm/shmem-helper: Avoid vm_open error paths

[Rob Clark]

From: Rob Clark <robdclark@chromium.org>

vm_open() is not allowed to fail.  Fortunately we are guaranteed that
the pages are already pinned, thanks to the initial mmap which is now
being cloned into a forked process, and only need to increment the
refcnt.  So just increment it directly.  Previously if a signal was
delivered at the wrong time to the forking process, the
mutex_lock_interruptible() could fail resulting in the pages_use_count
not being incremented.

Fixes: 2194a63a818d ("drm: Add library for shmem backed GEM objects")
Cc: stable@vger.kernel.org
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/drm_gem_shmem_helper.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem_shmem_helper.c b/drivers/gpu/drm/drm_gem_shmem_helper.c
index 3b7b71391a4c..b602cd72a120 100644
--- a/drivers/gpu/drm/drm_gem_shmem_helper.c
+++ b/drivers/gpu/drm/drm_gem_shmem_helper.c
@@ -571,12 +571,20 @@ static void drm_gem_shmem_vm_open(struct vm_area_struct *vma)
 {
 	struct drm_gem_object *obj = vma->vm_private_data;
 	struct drm_gem_shmem_object *shmem = to_drm_gem_shmem_obj(obj);
-	int ret;
 
 	WARN_ON(shmem->base.import_attach);
 
-	ret = drm_gem_shmem_get_pages(shmem);
-	WARN_ON_ONCE(ret != 0);
+	mutex_lock(&shmem->pages_lock);
+
+	/*
+	 * We should have already pinned the pages when the buffer was first
+	 * mmap'd, vm_open() just grabs an additional reference for the new
+	 * mm the vma is getting copied into (ie. on fork()).
+	 */
+	if (!WARN_ON_ONCE(!shmem->pages_use_count))
+		shmem->pages_use_count++;
+
+	mutex_unlock(&shmem->pages_lock);
 
 	drm_gem_vm_open(vma);
 }
-- 
2.38.1

Re: [PATCH v2 1/2] drm/shmem-helper: Remove errant put in error path

[Dmitry Osipenko]

On 11/30/22 21:57, Rob Clark wrote:
> From: Rob Clark <robdclark@chromium.org>
> 
> drm_gem_shmem_mmap() doesn't own this reference, resulting in the GEM
> object getting prematurely freed leading to a later use-after-free.
> 
> Link: https://syzkaller.appspot.com/bug?extid=c8ae65286134dd1b800d
> Reported-by: syzbot+c8ae65286134dd1b800d@syzkaller.appspotmail.com
> Fixes: 2194a63a818d ("drm: Add library for shmem backed GEM objects")
> Cc: stable@vger.kernel.org
> Signed-off-by: Rob Clark <robdclark@chromium.org>
> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> ---
>  drivers/gpu/drm/drm_gem_shmem_helper.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_gem_shmem_helper.c b/drivers/gpu/drm/drm_gem_shmem_helper.c
> index 35138f8a375c..3b7b71391a4c 100644
> --- a/drivers/gpu/drm/drm_gem_shmem_helper.c
> +++ b/drivers/gpu/drm/drm_gem_shmem_helper.c
> @@ -622,10 +622,8 @@ int drm_gem_shmem_mmap(struct drm_gem_shmem_object *shmem, struct vm_area_struct
>  	}
>  
>  	ret = drm_gem_shmem_get_pages(shmem);
> -	if (ret) {
> -		drm_gem_vm_close(vma);
> +	if (ret)
>  		return ret;
> -	}
>  
>  	vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
>  	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);

AFAICS, the dmabuf mmaping code path needs a similar fix, isn't it?

-		/* Drop the reference drm_gem_mmap_obj() acquired.*/
-		drm_gem_object_put(obj);
		vma->vm_private_data = NULL;

-		return dma_buf_mmap(obj->dma_buf, vma, 0);
+		ret = dma_buf_mmap(obj->dma_buf, vma, 0);
+
+		/* Drop the reference drm_gem_mmap_obj() acquired.*/
+		if (!ret)
+			drm_gem_object_put(obj);
+
+		return ret;


-- 
Best regards,
Dmitry

Re: [PATCH v2 1/2] drm/shmem-helper: Remove errant put in error path

[Rob Clark]

On Sun, Dec 4, 2022 at 12:45 PM Dmitry Osipenko
<dmitry.osipenko@collabora.com> wrote:
>
> On 11/30/22 21:57, Rob Clark wrote:
> > From: Rob Clark <robdclark@chromium.org>
> >
> > drm_gem_shmem_mmap() doesn't own this reference, resulting in the GEM
> > object getting prematurely freed leading to a later use-after-free.
> >
> > Link: https://syzkaller.appspot.com/bug?extid=c8ae65286134dd1b800d
> > Reported-by: syzbot+c8ae65286134dd1b800d@syzkaller.appspotmail.com
> > Fixes: 2194a63a818d ("drm: Add library for shmem backed GEM objects")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Rob Clark <robdclark@chromium.org>
> > Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> > ---
> >  drivers/gpu/drm/drm_gem_shmem_helper.c | 4 +---
> >  1 file changed, 1 insertion(+), 3 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/drm_gem_shmem_helper.c b/drivers/gpu/drm/drm_gem_shmem_helper.c
> > index 35138f8a375c..3b7b71391a4c 100644
> > --- a/drivers/gpu/drm/drm_gem_shmem_helper.c
> > +++ b/drivers/gpu/drm/drm_gem_shmem_helper.c
> > @@ -622,10 +622,8 @@ int drm_gem_shmem_mmap(struct drm_gem_shmem_object *shmem, struct vm_area_struct
> >       }
> >
> >       ret = drm_gem_shmem_get_pages(shmem);
> > -     if (ret) {
> > -             drm_gem_vm_close(vma);
> > +     if (ret)
> >               return ret;
> > -     }
> >
> >       vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
> >       vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
>
> AFAICS, the dmabuf mmaping code path needs a similar fix, isn't it?
>
> -               /* Drop the reference drm_gem_mmap_obj() acquired.*/
> -               drm_gem_object_put(obj);
>                 vma->vm_private_data = NULL;
>
> -               return dma_buf_mmap(obj->dma_buf, vma, 0);
> +               ret = dma_buf_mmap(obj->dma_buf, vma, 0);
> +
> +               /* Drop the reference drm_gem_mmap_obj() acquired.*/
> +               if (!ret)
> +                       drm_gem_object_put(obj);
> +
> +               return ret;
>

Yes, it seems that way.. I wish the shmem helpers worked in a less
"special" way with regards to refcnting :-(

BR,
-R