Am 10.11.2017 um 08:22 schrieb Chunming Zhou:
>
>
> On 2017å¹´11æ??09æ?¥ 16:59, Christian König wrote:
>> Deleted BOs with the same reservation object can be reaped even if they
>> can't be reserved.
>>
>> v2: rebase and we still need to remove/add the BO from/to the LRU.
>>
>> Signed-off-by: Christian König <christian.koenig at amd.com>
>> ---
>> Â drivers/gpu/drm/ttm/ttm_bo.c | 39
>> +++++++++++++++++++++++++++++++--------
>> Â 1 file changed, 31 insertions(+), 8 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
>> index 50a678b504f3..6545c4344684 100644
>> --- a/drivers/gpu/drm/ttm/ttm_bo.c
>> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
>> @@ -735,20 +735,37 @@ bool ttm_bo_eviction_valuable(struct
>> ttm_buffer_object *bo,
>> Â EXPORT_SYMBOL(ttm_bo_eviction_valuable);
>> Â Â static int ttm_mem_evict_first(struct ttm_bo_device *bdev,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â uint32_t mem_type,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â const struct ttm_place *place,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â bool interruptible,
>> -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â bool no_wait_gpu)
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â struct reservation_object *resv,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â uint32_t mem_type,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â const struct ttm_place *place,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â bool interruptible,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â bool no_wait_gpu)
>> Â {
>> Â Â Â Â Â struct ttm_bo_global *glob = bdev->glob;
>> Â Â Â Â Â struct ttm_mem_type_manager *man = &bdev->man[mem_type];
>> Â Â Â Â Â struct ttm_buffer_object *bo;
>> Â Â Â Â Â int ret = -EBUSY;
>> +Â Â Â bool locked;
>> Â Â Â Â Â unsigned i;
>> Â Â Â Â Â Â spin_lock(&glob->lru_lock);
>> Â Â Â Â Â for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
>> Â Â Â Â Â Â Â Â Â list_for_each_entry(bo, &man->lru[i], lru) {
>> +Â Â Â Â Â Â Â Â Â Â Â if (bo->resv == resv) {
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â if (list_empty(&bo->ddestroy))
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â continue;
> I don't think only destroying BO can be evicted under per-vm-bo case,
> but also normal BO should as well.
> I'd give an example:
> 1. vm-A allocates all vram;
> 2. vm-B also try to allocate full vram, so the BOs of vram in vm-A
> will be evicted to GTT.
> 3. vm-A is trying to allocate all GTT, if we don't allow eviction or
> swap, then will fail here.
That is a really good example, thanks.
>
> As above, we shouldn't disallow eviction and swap during allocation,
> we aren't able to predict what case happen.
> For over limit allocation, at worst, they will be returned with failed
> status while doing its CS.
> If you think the allocation shouldn't be over limitation of memory, we
> can add the checking condition before allocation every time, but not
> disallow eviction and swap in allocation, which really breaks the used
> TTM.
Ok, you convinced me. The case above indeed needs a better handling.
I will reactivate my operation context patch set for TTM. Shouldn't be
to much work to get this going.
Regards,
Christian.
>
> Regards,
> David Zhou
>
>> +
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â if (place &&
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â !bdev->driver->eviction_valuable(bo, place))
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â continue;
>> +
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â ttm_bo_del_from_lru(bo);
>> +
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â ret = 0;
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â locked = false;
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
>> +Â Â Â Â Â Â Â Â Â Â Â }
>> +
>> Â Â Â Â Â Â Â Â Â Â Â Â Â ret = reservation_object_trylock(bo->resv) ? 0 : -EBUSY;
>> Â Â Â Â Â Â Â Â Â Â Â Â Â if (ret)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â continue;
>> @@ -760,6 +777,7 @@ static int ttm_mem_evict_first(struct
>> ttm_bo_device *bdev,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â continue;
>> Â Â Â Â Â Â Â Â Â Â Â Â Â }
>> Â +Â Â Â Â Â Â Â Â Â Â Â locked = true;
>> Â Â Â Â Â Â Â Â Â Â Â Â Â break;
>> Â Â Â Â Â Â Â Â Â }
>> Â @@ -775,7 +793,8 @@ static int ttm_mem_evict_first(struct
>> ttm_bo_device *bdev,
>> Â Â Â Â Â kref_get(&bo->list_kref);
>> Â Â Â Â Â Â if (!list_empty(&bo->ddestroy)) {
>> -Â Â Â Â Â Â Â ret = ttm_bo_cleanup_refs(bo, interruptible, no_wait_gpu,
>> true);
>> +Â Â Â Â Â Â Â ret = ttm_bo_cleanup_refs(bo, interruptible, no_wait_gpu,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â locked);
>> Â Â Â Â Â Â Â Â Â kref_put(&bo->list_kref, ttm_bo_release_list);
>> Â Â Â Â Â Â Â Â Â return ret;
>> Â Â Â Â Â }
>> @@ -786,7 +805,10 @@ static int ttm_mem_evict_first(struct
>> ttm_bo_device *bdev,
>> Â Â Â Â Â BUG_ON(ret != 0);
>> Â Â Â Â Â Â ret = ttm_bo_evict(bo, interruptible, no_wait_gpu);
>> -Â Â Â ttm_bo_unreserve(bo);
>> +Â Â Â if (locked)
>> +Â Â Â Â Â Â Â ttm_bo_unreserve(bo);
>> +Â Â Â else
>> +Â Â Â Â Â Â Â ttm_bo_add_to_lru(bo);
>> Â Â Â Â Â Â kref_put(&bo->list_kref, ttm_bo_release_list);
>> Â Â Â Â Â return ret;
>> @@ -850,7 +872,7 @@ static int ttm_bo_mem_force_space(struct
>> ttm_buffer_object *bo,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â return ret;
>> Â Â Â Â Â Â Â Â Â if (mem->mm_node)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â break;
>> -Â Â Â Â Â Â Â ret = ttm_mem_evict_first(bdev, mem_type, place,
>> +Â Â Â Â Â Â Â ret = ttm_mem_evict_first(bdev, bo->resv, mem_type, place,
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â interruptible, no_wait_gpu);
>> Â Â Â Â Â Â Â Â Â if (unlikely(ret != 0))
>> Â Â Â Â Â Â Â Â Â Â Â Â Â return ret;
>> @@ -1353,7 +1375,8 @@ static int ttm_bo_force_list_clean(struct
>> ttm_bo_device *bdev,
>> Â Â Â Â Â for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
>> Â Â Â Â Â Â Â Â Â while (!list_empty(&man->lru[i])) {
>> Â Â Â Â Â Â Â Â Â Â Â Â Â spin_unlock(&glob->lru_lock);
>> -Â Â Â Â Â Â Â Â Â Â Â ret = ttm_mem_evict_first(bdev, mem_type, NULL, false,
>> false);
>> +Â Â Â Â Â Â Â Â Â Â Â ret = ttm_mem_evict_first(bdev, NULL, mem_type, NULL,
>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â false, false);
>> Â Â Â Â Â Â Â Â Â Â Â Â Â if (ret)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â return ret;
>> Â Â Â Â Â Â Â Â Â Â Â Â Â spin_lock(&glob->lru_lock);
>
