Am 10.04.2018 um 20:25 schrieb Sinan Kaya: > Code is expecing to observe the same number of buffers returned from > dma_map_sg() function compared to sg_alloc_table_from_pages(). This > doesn't hold true universally especially for systems with IOMMU. > > IOMMU driver tries to combine buffers into a single DMA address as much > as it can. The right thing is to tell the DMA layer how much combining > IOMMU can do. Good catch, but wrong place to set this. Please move it into the device initialization functions. Regards, Christian. > > Signed-off-by: Sinan Kaya <okaya at codeaurora.org> > --- > drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c > index e4bb435..02465cd 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c > @@ -787,6 +787,8 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) > enum dma_data_direction direction = write ? > DMA_BIDIRECTIONAL : DMA_TO_DEVICE; > > + dma_set_max_seg_size(adev->dev, PAGE_SIZE); > + > r = sg_alloc_table_from_pages(ttm->sg, ttm->pages, ttm->num_pages, 0, > ttm->num_pages << PAGE_SHIFT, > GFP_KERNEL);