Hi Robin:

On 6/28/21 5:54 PM, Robin Murphy wrote:

On 2021-06-27 09:47, Andy Yan wrote:

...

When iommu itself is disabled in dts, we should fallback to non-iommu buffer, check iommu parent is meanless here.

Signed-off-by: Andy Yan andy.yan@rock-chips.com

drivers/gpu/drm/rockchip/rockchip_drm_drv.c | 2 +-   1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c index 212bd87c0c4a..0d20c8cc1ffa 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.c @@ -366,7 +366,7 @@ static int rockchip_drm_platform_of_probe(struct device *dev)           }             iommu = of_parse_phandle(port->parent, "iommus", 0); -        if (!iommu || !of_device_is_available(iommu->parent)) { +        if (!iommu || !of_device_is_available(iommu)) {

Just wondering, could you use device_iommu_mapped() to avoid poking at the DT at all here, or is it that you need to check remote graph endpoints whose struct device may not be ready yet?

device_iommu_mapped(dev)  can't be called from rockchip_drm_init_iommu, as there is no  iommu dev(vop) in these function.

Robin.

...

DRM_DEV_DEBUG(dev,                         "no iommu attached for %pOF, using non-iommu buffers\n",                         port->parent);