Re: [Intel-gfx] [PATCH v3 06/12] drm/ttm: add TTM_PAGE_FLAG_EXTERNAL_MAPPABLE

[Matthew Auld <matthew.auld@intel.com>]

On 16/09/2021 10:03, Thomas Hellström wrote:
> On Thu, 2021-09-16 at 08:55 +0200, Christian König wrote:
>>
>>
>> Am 15.09.21 um 20:59 schrieb Matthew Auld:
>>> In commit:
>>>
>>> commit 667a50db0477d47fdff01c666f5ee1ce26b5264c
>>> Author: Thomas Hellstrom <thellstrom@vmware.com>
>>> Date:   Fri Jan 3 11:17:18 2014 +0100
>>>
>>>       drm/ttm: Refuse to fault (prime-) imported pages
>>>
>>> we introduced the restriction that imported pages should not be
>>> directly
>>> mappable through TTM(this also extends to userptr). In the next
>>> patch we
>>> want to introduce a shmem_tt backend, which should follow all the
>>> existing rules with TTM_PAGE_FLAG_EXTERNAL, since it will need to
>>> handle
>>> swapping itself, but with the above mapping restriction lifted.
>>>
>>> Signed-off-by: Matthew Auld <matthew.auld@intel.com>
>>> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
>>> Cc: Christian König <christian.koenig@amd.com>
>>> ---
>>>    drivers/gpu/drm/ttm/ttm_bo_vm.c | 6 ++++--
>>>    include/drm/ttm/ttm_tt.h        | 7 +++++++
>>>    2 files changed, 11 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c
>>> b/drivers/gpu/drm/ttm/ttm_bo_vm.c
>>> index 708390588c7c..fd6e18f12f50 100644
>>> --- a/drivers/gpu/drm/ttm/ttm_bo_vm.c
>>> +++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c
>>> @@ -163,8 +163,10 @@ vm_fault_t ttm_bo_vm_reserve(struct
>>> ttm_buffer_object *bo,
>>>           * (if at all) by redirecting mmap to the exporter.
>>>           */
>>>          if (bo->ttm && (bo->ttm->page_flags &
>>> TTM_PAGE_FLAG_EXTERNAL)) {
>>> -               dma_resv_unlock(bo->base.resv);
>>> -               return VM_FAULT_SIGBUS;
>>> +               if (!(bo->ttm->page_flags &
>>> TTM_PAGE_FLAG_EXTERNAL_MAPPABLE)) {
> 
> I saw there was previously a thought about testing for ttm_bo_type_sg
> here. Was that a dead end?

I guess the issue was userptr, which is also EXTERNAL and needs to be 
rejected here, but underneath it's not type_sg.

Just having an explicit MAPPABLE flag or similar seemed the more 
flexible approach? Happy to try out other ideas.

> 
> Thanks,
> Thomas
> 
>