drm/amdgpu: fix scatter-gather mapping with user pages

Calls to dma_map_sg may return less segments / entries than requested
if they fall on page bounderies. The old implementation did not
support this use case.

Fixes: be62dbf ("iommu/amd: Convert AMD iommu driver to the dma-iommu api")
Bug: https://bugzilla.kernel.org/show_bug.cgi?id=206461
Bug: https://bugzilla.kernel.org/show_bug.cgi?id=206895
Bug: https://gitlab.freedesktop.org/drm/amd/issues/1056
Signed-off-by: Shane Francis <[email protected]>
Reviewed-by: Michael J. Ruhl <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Cc: [email protected]

Author: BuzzBumbleBee

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -974,7 +974,7 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
 	/* Map SG to device */
 	r = -ENOMEM;
 	nents = dma_map_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction);
-	if (nents != ttm->sg->nents)
+	if (nents == 0)
 		goto release_sg;
 
 	/* convert SG to linear array of pages and dma addresses */