Re: [PATCH] drm/amdkfd: Fix potential deallocation of previously deallocated memory.

[Andi Shyti <andi.shyti@linux.intel.com>]

On Tue, Apr 18, 2023 at 12:07:15PM +0200, Krzysztof Kozlowski wrote:
> On 18/04/2023 10:47, Andi Shyti wrote:
> > Hi Daniil,
> > 
> > On Mon, Apr 17, 2023 at 11:55:21PM -0700, Daniil Dulov wrote:
> >> Pointer mqd_mem_obj can be deallocated in kfd_gtt_sa_allocate().
> >> The function then returns non-zero value, which causes the second deallocation.
> >>
> >> Found by Linux Verification Center (linuxtesting.org) with SVACE.
> >>
> >> Fixes: d1f8f0d17d40 ("drm/amdkfd: Move non-sdma mqd allocation out of init_mqd")
> >> Signed-off-by: Daniil Dulov <d.dulov@aladdin.ru>
> >> ---
> >>  drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c | 3 ++-
> >>  1 file changed, 2 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
> >> index 3b6f5963180d..bce11c5b07d6 100644
> >> --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
> >> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
> >> @@ -119,7 +119,8 @@ static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
> >>  	}
> >>  
> >>  	if (retval) {
> >> -		kfree(mqd_mem_obj);
> >> +		if (mqd_mem_obj)
> >> +			kfree(mqd_mem_obj);
> > 
> > I think this is not needed. kfree() returns immediately if
> > mqd_mem_obj is NULL.
> > 
> 
> Yep, the tool has to be fixed because such patch is just misleading.
> However different point - the commit description actually describes
> entirely different case: double free. Maybe the issue is true, just the
> fix is wrong?

Yes, indeed, the fix is wrong, but the bug exists. I'm pasting
the original function:

	if (kfd->cwsr_enabled && (q->type == KFD_QUEUE_TYPE_COMPUTE)) {
		mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL);
		if (!mqd_mem_obj)
			return NULL;
		...
	} else {
		retval = kfd_gtt_sa_allocate(kfd, sizeof(struct v9_mqd),
				&mqd_mem_obj);
	}

	if (retval) {
		kfree(mqd_mem_obj);
		return NULL;
	}

The "kfd_gtt_sa_allocate()" function allocates mqd_mem_obj and if
an error occurs internally frees it, without setting it to NULL;
retval is true and we kfree a memory that has already been freed.

The real fix is to move the "if (retval)" inside the if. It would
basically be:

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index fdbfd725841ff..31d47d687bd62 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -115,18 +115,20 @@ static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
                        &(mqd_mem_obj->gtt_mem),
                        &(mqd_mem_obj->gpu_addr),
                        (void *)&(mqd_mem_obj->cpu_ptr), true);
+
+               if (retval) {
+                       kfree(mqd_mem_obj);
+                       return NULL;
+               }
+
        } else {
                retval = kfd_gtt_sa_allocate(kfd, sizeof(struct v9_mqd),
                                &mqd_mem_obj);
-       }
-
-       if (retval) {
-               kfree(mqd_mem_obj);
-               return NULL;
+               if (retval)
+                       return NULL;
        }
 
        return mqd_mem_obj;
-
 }

Maybe with some clever refactoring we could reduce some code
duplication.

Daniil will you look into this?

Andi