Nouveau - Feb 2018 - [drm-nouveau-mmu] question about potential NULL pointer dereference

Hi all,

While doing some static analysis I ran into the following piece of  
code at drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.c:957:

  957#define node(root, dir) ((root)->head.dir == &vmm->list) ? NULL :
              \
  958        list_entry((root)->head.dir, struct nvkm_vma, head)
  959
  960void
  961nvkm_vmm_unmap_region(struct nvkm_vmm *vmm, struct nvkm_vma *vma)
  962{
  963        struct nvkm_vma *next;
  964
  965        nvkm_memory_tags_put(vma->memory,  
vmm->mmu->subdev.device, &vma->tags);
  966        nvkm_memory_unref(&vma->memory);
  967
  968        if (vma->part) {
  969                struct nvkm_vma *prev = node(vma, prev);
  970                if (!prev->memory) {
  971                        prev->size += vma->size;
  972                        rb_erase(&vma->tree, &vmm->root);
  973                        list_del(&vma->head);
  974                        kfree(vma);
  975                        vma = prev;
  976                }
  977        }
  978
  979        next = node(vma, next);
  980        if (next && next->part) {
  981                if (!next->memory) {
  982                        vma->size += next->size;
  983                        rb_erase(&next->tree, &vmm->root);
  984                        list_del(&next->head);
  985                        kfree(next);
  986                }
  987        }
  988}

The issue here is that in case _node_ returns NULL, _prev_ is not  
being null checked, hence there is a potential null pointer  
dereference at line 970.

Notice that _next_ is being null checked at line 980, so I wonder if  
_prev_ should be checked the same as _next_.

The fact that both _next_ and next->part are null checked, makes me  
wonder if in case _prev_ actually needs to be checked, there is  
another pointer contained into _prev_ to be validated as well? I'm  
sorry, this is not clear to me at this moment.

I appreciate your feedback
Thank you
--
Gustavo

On Wed, Feb 14, 2018 at 1:40 AM, Gustavo A. R. Silva
<garsilva at embeddedor.com> wrote:
>
> Hi all,
>
> While doing some static analysis I ran into the following piece of code at
> drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.c:957:
>
>  957#define node(root, dir) ((root)->head.dir == &vmm->list) ?
NULL :
> \
>  958        list_entry((root)->head.dir, struct nvkm_vma, head)
>  959
>  960void
>  961nvkm_vmm_unmap_region(struct nvkm_vmm *vmm, struct nvkm_vma *vma)
>  962{
>  963        struct nvkm_vma *next;
>  964
>  965        nvkm_memory_tags_put(vma->memory,
vmm->mmu->subdev.device,
> &vma->tags);
>  966        nvkm_memory_unref(&vma->memory);
>  967
>  968        if (vma->part) {
>  969                struct nvkm_vma *prev = node(vma, prev);
>  970                if (!prev->memory) {
>  971                        prev->size += vma->size;
>  972                        rb_erase(&vma->tree, &vmm->root);
>  973                        list_del(&vma->head);
>  974                        kfree(vma);
>  975                        vma = prev;
>  976                }
>  977        }
>  978
>  979        next = node(vma, next);
>  980        if (next && next->part) {
>  981                if (!next->memory) {
>  982                        vma->size += next->size;
>  983                        rb_erase(&next->tree,
&vmm->root);
>  984                        list_del(&next->head);
>  985                        kfree(next);
>  986                }
>  987        }
>  988}
>
> The issue here is that in case _node_ returns NULL, _prev_ is not being
null
> checked, hence there is a potential null pointer dereference at line 970.
>
> Notice that _next_ is being null checked at line 980, so I wonder if _prev_
> should be checked the same as _next_.
>
> The fact that both _next_ and next->part are null checked, makes me
wonder
> if in case _prev_ actually needs to be checked, there is another pointer
> contained into _prev_ to be validated as well? I'm sorry, this is not
clear
> to me at this moment.
It's not checked because it can't happen.  If vma->part is set, there
will be a previous node that it was split from.

Ben.
>
> I appreciate your feedback
> Thank you
> --
> Gustavo
>
>
>
>
>
>

Quoting Ben Skeggs <bskeggs at redhat.com>:
> On Wed, Feb 14, 2018 at 1:40 AM, Gustavo A. R. Silva
> <garsilva at embeddedor.com> wrote:
>>
>> Hi all,
>>
>> While doing some static analysis I ran into the following piece of code
at
>> drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.c:957:
>>
>>  957#define node(root, dir) ((root)->head.dir == &vmm->list)
? NULL :
>> \
>>  958        list_entry((root)->head.dir, struct nvkm_vma, head)
>>  959
>>  960void
>>  961nvkm_vmm_unmap_region(struct nvkm_vmm *vmm, struct nvkm_vma *vma)
>>  962{
>>  963        struct nvkm_vma *next;
>>  964
>>  965        nvkm_memory_tags_put(vma->memory,
vmm->mmu->subdev.device,
>> &vma->tags);
>>  966        nvkm_memory_unref(&vma->memory);
>>  967
>>  968        if (vma->part) {
>>  969                struct nvkm_vma *prev = node(vma, prev);
>>  970                if (!prev->memory) {
>>  971                        prev->size += vma->size;
>>  972                        rb_erase(&vma->tree,
&vmm->root);
>>  973                        list_del(&vma->head);
>>  974                        kfree(vma);
>>  975                        vma = prev;
>>  976                }
>>  977        }
>>  978
>>  979        next = node(vma, next);
>>  980        if (next && next->part) {
>>  981                if (!next->memory) {
>>  982                        vma->size += next->size;
>>  983                        rb_erase(&next->tree,
&vmm->root);
>>  984                        list_del(&next->head);
>>  985                        kfree(next);
>>  986                }
>>  987        }
>>  988}
>>
>> The issue here is that in case _node_ returns NULL, _prev_ is not being
null
>> checked, hence there is a potential null pointer dereference at line
970.
>>
>> Notice that _next_ is being null checked at line 980, so I wonder if
_prev_
>> should be checked the same as _next_.
>>
>> The fact that both _next_ and next->part are null checked, makes me
wonder
>> if in case _prev_ actually needs to be checked, there is another
pointer
>> contained into _prev_ to be validated as well? I'm sorry, this is
not clear
>> to me at this moment.
> It's not checked because it can't happen.  If vma->part is set,
there
> will be a previous node that it was split from.
>
I got it.

Thanks, Ben.
--
Gustavo