Hello,
On 2016-02-12 06:55, Inki Dae wrote:
> Hi Marek,
>
> 2016ë
02ì 03ì¼ 21:42ì Marek Szyprowski ì´(ê°) ì´ ê¸:
>> Virtual DRM device is used for memory allocation for GEM objects. On
>> some architectures null dma operations are assigned for virtual devices
>> if no explicit initialization has been made.
>>
>> Signed-off-by: Marek Szyprowski <m.szyprowski at samsung.com>
>> ---
>> drivers/gpu/drm/exynos/exynos_drm_drv.c | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c
>> b/drivers/gpu/drm/exynos/exynos_drm_drv.c
>> index 68f0f36..a33ea44 100644
>> --- a/drivers/gpu/drm/exynos/exynos_drm_drv.c
>> +++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c
>> @@ -144,6 +144,7 @@ static int exynos_drm_load(struct drm_device *dev,
>> unsigned long flags)
>> init_waitqueue_head(&private->wait);
>> spin_lock_init(&private->lock);
>>
>> + arch_setup_dma_ops(dev->dev, 0, DMA_BIT_MASK(32), false, false);
> As you got the report, above function incurrs build error below,
> ERROR: "arch_setup_dma_ops" [drivers/gpu/drm/exynos/exynosdrm.ko] undefined!
>
> Exynos drm can be built as a module. In this case, arch_setup_dma_ops cannot
> be accessed by the module.
>
> I could fix it easily like below but I'm not sure that it's reasonable.
>
> +extern void arch_setup_dma_ops(struct device *dev, u64 dma_base,
> + u64 size, struct iommu_ops *iommu,
> + bool coherent) { };
I don't think this is a good idea - the result will be same as not applying
the patch at all. Please drop this patch for now, I will rework it while
adding support for generic IOMMU for ARM and ARM64.
Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland
