sw64: gpu: correct low-level mmio memset/memcpy direct calls

Sunway inclusion
category: bugfix
bugzilla: https://gitee.com/openeuler/kernel/issues/I5GDKC

--------------------------------

Driver codes of the direct calls, via the SIMD-optimized memset
and memcpy functions, may raise DFAULT when using RX580 or R7
Series graphics card on sw64, so work around 'memset' references
to '_memset_c_io' and 'memcpy' to 'memcpy_fromio'.
Signed-off-by: NXiong Aifei <xiongaifei@wxiat.com>
Signed-off-by: NGu Zitao <guzitao@wxiat.com>

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

@@ -1354,7 +1354,11 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev)
 			return r;
 		}
 
+#if IS_ENABLED(CONFIG_SW64)
+		_memset_c_io(hpd, 0, mec_hpd_size);
+#else
 		memset(hpd, 0, mec_hpd_size);
+#endif
 
 		amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj);
 		amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj);
@@ -4649,7 +4653,11 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 		vi_srbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	} else {
+#if IS_ENABLED(CONFIG_SW64)
+		_memset_c_io((void *)mqd, 0, sizeof(struct vi_mqd_allocation));
+#else
 		memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation));
+#endif
 		((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
@@ -4660,7 +4668,11 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
+#if IS_ENABLED(CONFIG_SW64)
+			memcpy_fromio(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
+#else
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
+#endif
 	}
 
 	return 0;

drivers/gpu/drm/radeon/vce_v1_0.c

@@ -193,8 +193,13 @@ int vce_v1_0_load_fw(struct radeon_device *rdev, uint32_t *data)
 	data[3] = sign->val[i].nonce[3];
 	data[4] = cpu_to_le32(le32_to_cpu(sign->len) + 64);
 
+#if IS_ENABLED(CONFIG_SW64)
+	memset_io(&data[5], 0, 44);
+	memcpy_toio(&data[16], &sign[1], rdev->vce_fw->size - sizeof(*sign));
+#else
 	memset(&data[5], 0, 44);
 	memcpy(&data[16], &sign[1], rdev->vce_fw->size - sizeof(*sign));
+#endif
 
 	data += (le32_to_cpu(sign->len) + 64) / 4;
 	data[0] = sign->val[i].sigval[0];