[PATCH v2 11/11] drm/amdgpu: set TMZ bits in PTEs for secure BO (v4)

[Christian König]

Am 25.09.19 um 18:01 schrieb Alex Deucher:
> On Wed, Sep 25, 2019 at 9:59 AM Koenig, Christian
> <Christian.Koenig at amd.com> wrote:
>> Am 25.09.19 um 15:45 schrieb Huang, Ray:
>>> From: Alex Deucher <alexander.deucher at amd.com>
>>>
>>> If a buffer object is secure, i.e. created with
>>> AMDGPU_GEM_CREATE_ENCRYPTED, then the TMZ bit of
>>> the PTEs that belong the buffer object should be
>>> set.
>>>
>>> v1: design and draft the skeletion of TMZ bits setting on PTEs (Alex)
>>> v2: return failure once create secure BO on non-TMZ platform  (Ray)
>>> v3: amdgpu_bo_encrypted() only checks the BO (Luben)
>>> v4: move TMZ flag setting into amdgpu_vm_bo_update  (Christian)
>>>
>>> Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
>>> Reviewed-by: Huang Rui <ray.huang at amd.com>
>>> Signed-off-by: Huang Rui <ray.huang at amd.com>
>>> Signed-off-by: Luben Tuikov <luben.tuikov at amd.com>
>>> Reviewed-by: Alex Deucher <alexander.deucher at amd.com>
>>> ---
>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c    | 12 +++++++++++-
>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 11 +++++++++++
>>>    drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c     |  5 +++++
>>>    3 files changed, 27 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>>> index 22eab74..5332104 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>>> @@ -222,7 +222,8 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
>>>                      AMDGPU_GEM_CREATE_CPU_GTT_USWC |
>>>                      AMDGPU_GEM_CREATE_VRAM_CLEARED |
>>>                      AMDGPU_GEM_CREATE_VM_ALWAYS_VALID |
>>> -                   AMDGPU_GEM_CREATE_EXPLICIT_SYNC))
>>> +                   AMDGPU_GEM_CREATE_EXPLICIT_SYNC |
>>> +                   AMDGPU_GEM_CREATE_ENCRYPTED))
>>>
>>>                return -EINVAL;
>>>
>>> @@ -230,6 +231,11 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
>>>        if (args->in.domains & ~AMDGPU_GEM_DOMAIN_MASK)
>>>                return -EINVAL;
>>>
>>> +     if (!adev->tmz.enabled && (flags & AMDGPU_GEM_CREATE_ENCRYPTED)) {
>>> +             DRM_ERROR("Cannot allocate secure buffer while tmz is disabled\n");
>>> +             return -EINVAL;
>>> +     }
>>> +
>>>        /* create a gem object to contain this object in */
>>>        if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS |
>>>            AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) {
>>> @@ -251,6 +257,10 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
>>>                resv = vm->root.base.bo->tbo.resv;
>>>        }
>>>
>>> +     if (flags & AMDGPU_GEM_CREATE_ENCRYPTED) {
>>> +             /* XXX: pad out alignment to meet TMZ requirements */
>>> +     }
>>> +
>>>        r = amdgpu_gem_object_create(adev, size, args->in.alignment,
>>>                                     (u32)(0xffffffff & args->in.domains),
>>>                                     flags, ttm_bo_type_device, resv, &gobj);
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>>> index 5a3c177..75c7392 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
>>> @@ -224,6 +224,17 @@ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo)
>>>        return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC;
>>>    }
>>>
>>> +/**
>>> + * amdgpu_bo_encrypted - test if the BO is encrypted
>>> + * @bo: pointer to a buffer object
>>> + *
>>> + * Return true if the buffer object is encrypted, false otherwise.
>>> + */
>>> +static inline bool amdgpu_bo_encrypted(struct amdgpu_bo *bo)
>>> +{
>>> +     return bo->flags & AMDGPU_GEM_CREATE_ENCRYPTED;
>>> +}
>>> +
>>>    bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
>>>    void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain);
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>>> index b285ab2..8e13b1fd3 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>>> @@ -1688,6 +1688,11 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
>>>
>>>        if (bo) {
>>>                flags = amdgpu_ttm_tt_pte_flags(adev, bo->tbo.ttm, mem);
>>> +
>>> +             if (amdgpu_bo_encrypted(bo)) {
>>> +                     flags |= AMDGPU_PTE_TMZ;
>>> +             }
>>> +
>> You can drop the {} here, apart from that the patch is Reviewed-by:
>> Christian König <christian.koenig at amd.com>.
>>
>>   From the design it would be indeed nicer to have that in
>> amdgpu_ttm_tt_pte_flags(), but we would need to make sure that this is
>> not called any more from binding a tt.
> Don't we need that for things like evictions of tmz buffers?  That
> needs special handling in general, which I guess we can address at
> that time.

Yeah, exactly. The problem is that we sometimes doesn't use bo->mem, but 
rather a separate memory descriptor to get the PTE infos.

I think we could remove the call in amdgpu_ttm_backend_bind() where this 
is relevant, because we now manage GART separately.

But that's a larger change and probably not something for this patch set.

Going to put it on my TODO list.

Christian.

>
> Alex
>
>> Regards,
>> Christian.
>>
>>>                bo_adev = amdgpu_ttm_adev(bo->tbo.bdev);
>>>        } else {
>>>                flags = 0x0;
>> _______________________________________________
>> amd-gfx mailing list
>> amd-gfx at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
> _______________________________________________
> amd-gfx mailing list
> amd-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx