Merge tag 'v4.7-rc6' into patchwork

Linux 4.7-rc6

* tag 'v4.7-rc6': (1245 commits)
  Linux 4.7-rc6
  ovl: warn instead of error if d_type is not supported
  MIPS: Fix possible corruption of cache mode by mprotect.
  locks: use file_inode()
  usb: dwc3: st: Use explicit reset_control_get_exclusive() API
  phy: phy-stih407-usb: Use explicit reset_control_get_exclusive() API
  phy: miphy28lp: Inform the reset framework that our reset line may be shared
  namespace: update event counter when umounting a deleted dentry
  9p: use file_dentry()
  lockd: unregister notifier blocks if the service fails to come up completely
  ACPI,PCI,IRQ: correct operator precedence
  fuse: serialize dirops by default
  drm/i915: Fix missing unlock on error in i915_ppgtt_info()
  powerpc: Initialise pci_io_base as early as possible
  mfd: da9053: Fix compiler warning message for uninitialised variable
  mfd: max77620: Fix FPS switch statements
  phy: phy-stih407-usb: Inform the reset framework that our reset line may be shared
  usb: dwc3: st: Inform the reset framework that our reset line may be shared
  usb: host: ehci-st: Inform the reset framework that our reset line may be shared
  usb: host: ohci-st: Inform the reset framework that our reset line may be shared
  ...

1 files changed, 20 insertions, 2 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index 518dca43b133..9dc4e24e31e7 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -66,6 +66,16 @@ MODULE_FIRMWARE("radeon/mullins_sdma1.bin");
 
 u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev);
 
+
+static void cik_sdma_free_microcode(struct amdgpu_device *adev)
+{
+	int i;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+			release_firmware(adev->sdma.instance[i].fw);
+			adev->sdma.instance[i].fw = NULL;
+	}
+}
+
 /*
  * sDMA - System DMA
  * Starting with CIK, the GPU has new asynchronous
@@ -419,6 +429,8 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 		/* Initialize the ring buffer's read and write pointers */
 		WREG32(mmSDMA0_GFX_RB_RPTR + sdma_offsets[i], 0);
 		WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], 0);
+		WREG32(mmSDMA0_GFX_IB_RPTR + sdma_offsets[i], 0);
+		WREG32(mmSDMA0_GFX_IB_OFFSET + sdma_offsets[i], 0);
 
 		/* set the wb address whether it's enabled or not */
 		WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i],
@@ -446,7 +458,12 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
 
 		ring->ready = true;
+	}
+
+	cik_sdma_enable(adev, true);
 
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		ring = &adev->sdma.instance[i].ring;
 		r = amdgpu_ring_test_ring(ring);
 		if (r) {
 			ring->ready = false;
@@ -529,8 +546,8 @@ static int cik_sdma_start(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	/* unhalt the MEs */
-	cik_sdma_enable(adev, true);
+	/* halt the engine before programing */
+	cik_sdma_enable(adev, false);
 
 	/* start the gfx rings and rlc compute queues */
 	r = cik_sdma_gfx_resume(adev);
@@ -998,6 +1015,7 @@ static int cik_sdma_sw_fini(void *handle)
 	for (i = 0; i < adev->sdma.num_instances; i++)
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
 
+	cik_sdma_free_microcode(adev);
 	return 0;
 }