1 files changed, 223 insertions, 212 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index c216e16826c9..9e8715b4739d 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -21,8 +21,10 @@
  *
  * Authors: Alex Deucher
  */
+
 #include <linux/firmware.h>
-#include <drm/drmP.h>
+#include <linux/module.h>
+
 #include "amdgpu.h"
 #include "amdgpu_ucode.h"
 #include "amdgpu_trace.h"
@@ -52,29 +54,27 @@ static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev);
-static int cik_sdma_soft_reset(void *handle);
-
-MODULE_FIRMWARE("radeon/bonaire_sdma.bin");
-MODULE_FIRMWARE("radeon/bonaire_sdma1.bin");
-MODULE_FIRMWARE("radeon/hawaii_sdma.bin");
-MODULE_FIRMWARE("radeon/hawaii_sdma1.bin");
-MODULE_FIRMWARE("radeon/kaveri_sdma.bin");
-MODULE_FIRMWARE("radeon/kaveri_sdma1.bin");
-MODULE_FIRMWARE("radeon/kabini_sdma.bin");
-MODULE_FIRMWARE("radeon/kabini_sdma1.bin");
-MODULE_FIRMWARE("radeon/mullins_sdma.bin");
-MODULE_FIRMWARE("radeon/mullins_sdma1.bin");
+static int cik_sdma_soft_reset(struct amdgpu_ip_block *ip_block);
 
 u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev);
 
+MODULE_FIRMWARE("amdgpu/bonaire_sdma.bin");
+MODULE_FIRMWARE("amdgpu/bonaire_sdma1.bin");
+MODULE_FIRMWARE("amdgpu/hawaii_sdma.bin");
+MODULE_FIRMWARE("amdgpu/hawaii_sdma1.bin");
+MODULE_FIRMWARE("amdgpu/kaveri_sdma.bin");
+MODULE_FIRMWARE("amdgpu/kaveri_sdma1.bin");
+MODULE_FIRMWARE("amdgpu/kabini_sdma.bin");
+MODULE_FIRMWARE("amdgpu/kabini_sdma1.bin");
+MODULE_FIRMWARE("amdgpu/mullins_sdma.bin");
+MODULE_FIRMWARE("amdgpu/mullins_sdma1.bin");
 
 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;
-	}
+
+	for (i = 0; i < adev->sdma.num_instances; i++)
+		amdgpu_ucode_release(&adev->sdma.instance[i].fw);
 }
 
 /*
@@ -106,7 +106,6 @@ static void cik_sdma_free_microcode(struct amdgpu_device *adev)
 static int cik_sdma_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err = 0, i;
 
 	DRM_DEBUG("\n");
@@ -132,21 +131,22 @@ static int cik_sdma_init_microcode(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		if (i == 0)
-			snprintf(fw_name, sizeof(fw_name), "radeon/%s_sdma.bin", chip_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma.bin", chip_name);
 		else
-			snprintf(fw_name, sizeof(fw_name), "radeon/%s_sdma1.bin", chip_name);
-		err = request_firmware(&adev->sdma.instance[i].fw, fw_name, adev->dev);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma1.bin", chip_name);
 		if (err)
 			goto out;
-		err = amdgpu_ucode_validate(adev->sdma.instance[i].fw);
 	}
 out:
 	if (err) {
-		pr_err("cik_sdma: Failed to load firmware \"%s\"\n", fw_name);
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			release_firmware(adev->sdma.instance[i].fw);
-			adev->sdma.instance[i].fw = NULL;
-		}
+		pr_err("cik_sdma: Failed to load firmware \"%s_sdma%s.bin\"\n",
+		       chip_name, i == 0 ? "" : "1");
+		for (i = 0; i < adev->sdma.num_instances; i++)
+			amdgpu_ucode_release(&adev->sdma.instance[i].fw);
 	}
 	return err;
 }
@@ -162,7 +162,7 @@ static uint64_t cik_sdma_ring_get_rptr(struct amdgpu_ring *ring)
 {
 	u32 rptr;
 
-	rptr = ring->adev->wb.wb[ring->rptr_offs];
+	rptr = *ring->rptr_cpu_addr;
 
 	return (rptr & 0x3fffc) >> 2;
 }
@@ -177,9 +177,8 @@ static uint64_t cik_sdma_ring_get_rptr(struct amdgpu_ring *ring)
 static uint64_t cik_sdma_ring_get_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	u32 me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1;
 
-	return (RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me]) & 0x3fffc) >> 2;
+	return (RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me]) & 0x3fffc) >> 2;
 }
 
 /**
@@ -192,15 +191,14 @@ static uint64_t cik_sdma_ring_get_wptr(struct amdgpu_ring *ring)
 static void cik_sdma_ring_set_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	u32 me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1;
 
-	WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me],
-		       	(lower_32_bits(ring->wptr) << 2) & 0x3fffc);
+	WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me],
+	       (ring->wptr << 2) & 0x3fffc);
 }
 
 static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
 {
-	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
+	struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
 	int i;
 
 	for (i = 0; i < count; i++)
@@ -215,18 +213,22 @@ static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
  * cik_sdma_ring_emit_ib - Schedule an IB on the DMA engine
  *
  * @ring: amdgpu ring pointer
+ * @job: job to retrive vmid from
  * @ib: IB object to schedule
+ * @flags: unused
  *
  * Schedule an IB in the DMA ring (CIK).
  */
 static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring,
+				  struct amdgpu_job *job,
 				  struct amdgpu_ib *ib,
-				  unsigned vm_id, bool ctx_switch)
+				  uint32_t flags)
 {
-	u32 extra_bits = vm_id & 0xf;
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	u32 extra_bits = vmid & 0xf;
 
 	/* IB packet must end on a 8 DW boundary */
-	cik_sdma_ring_insert_nop(ring, (12 - (lower_32_bits(ring->wptr) & 7)) % 8);
+	cik_sdma_ring_insert_nop(ring, (4 - lower_32_bits(ring->wptr)) & 7);
 
 	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_INDIRECT_BUFFER, 0, extra_bits));
 	amdgpu_ring_write(ring, ib->gpu_addr & 0xffffffe0); /* base must be 32 byte aligned */
@@ -248,7 +250,7 @@ static void cik_sdma_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 			  SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */
 	u32 ref_and_mask;
 
-	if (ring == &ring->adev->sdma.instance[0].ring)
+	if (ring->me == 0)
 		ref_and_mask = GPU_HDP_FLUSH_DONE__SDMA0_MASK;
 	else
 		ref_and_mask = GPU_HDP_FLUSH_DONE__SDMA1_MASK;
@@ -261,18 +263,13 @@ static void cik_sdma_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, (0xfff << 16) | 10); /* retry count, poll interval */
 }
 
-static void cik_sdma_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
-{
-	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
-	amdgpu_ring_write(ring, mmHDP_DEBUG0);
-	amdgpu_ring_write(ring, 1);
-}
-
 /**
  * cik_sdma_ring_emit_fence - emit a fence on the DMA ring
  *
  * @ring: amdgpu ring pointer
- * @fence: amdgpu fence object
+ * @addr: address
+ * @seq: sequence number
+ * @flags: fence related flags
  *
  * Add a DMA fence packet to the ring to write
  * the fence seq number and DMA trap packet to generate
@@ -310,23 +307,15 @@ static void cik_sdma_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq
  */
 static void cik_sdma_gfx_stop(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].ring;
-	struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].ring;
 	u32 rb_cntl;
 	int i;
 
-	if ((adev->mman.buffer_funcs_ring == sdma0) ||
-	    (adev->mman.buffer_funcs_ring == sdma1))
-		amdgpu_ttm_set_active_vram_size(adev, adev->mc.visible_vram_size);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]);
 		rb_cntl &= ~SDMA0_GFX_RB_CNTL__RB_ENABLE_MASK;
 		WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], 0);
 	}
-	sdma0->ready = false;
-	sdma1->ready = false;
 }
 
 /**
@@ -342,6 +331,63 @@ static void cik_sdma_rlc_stop(struct amdgpu_device *adev)
 }
 
 /**
+ * cik_ctx_switch_enable - stop the async dma engines context switch
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable/disable the DMA MEs context switch.
+ *
+ * Halt or unhalt the async dma engines context switch (VI).
+ */
+static void cik_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
+{
+	u32 f32_cntl, phase_quantum = 0;
+	int i;
+
+	if (amdgpu_sdma_phase_quantum) {
+		unsigned value = amdgpu_sdma_phase_quantum;
+		unsigned unit = 0;
+
+		while (value > (SDMA0_PHASE0_QUANTUM__VALUE_MASK >>
+				SDMA0_PHASE0_QUANTUM__VALUE__SHIFT)) {
+			value = (value + 1) >> 1;
+			unit++;
+		}
+		if (unit > (SDMA0_PHASE0_QUANTUM__UNIT_MASK >>
+			    SDMA0_PHASE0_QUANTUM__UNIT__SHIFT)) {
+			value = (SDMA0_PHASE0_QUANTUM__VALUE_MASK >>
+				 SDMA0_PHASE0_QUANTUM__VALUE__SHIFT);
+			unit = (SDMA0_PHASE0_QUANTUM__UNIT_MASK >>
+				SDMA0_PHASE0_QUANTUM__UNIT__SHIFT);
+			WARN_ONCE(1,
+			"clamping sdma_phase_quantum to %uK clock cycles\n",
+				  value << unit);
+		}
+		phase_quantum =
+			value << SDMA0_PHASE0_QUANTUM__VALUE__SHIFT |
+			unit  << SDMA0_PHASE0_QUANTUM__UNIT__SHIFT;
+	}
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		f32_cntl = RREG32(mmSDMA0_CNTL + sdma_offsets[i]);
+		if (enable) {
+			f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
+					AUTO_CTXSW_ENABLE, 1);
+			if (amdgpu_sdma_phase_quantum) {
+				WREG32(mmSDMA0_PHASE0_QUANTUM + sdma_offsets[i],
+				       phase_quantum);
+				WREG32(mmSDMA0_PHASE1_QUANTUM + sdma_offsets[i],
+				       phase_quantum);
+			}
+		} else {
+			f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
+					AUTO_CTXSW_ENABLE, 0);
+		}
+
+		WREG32(mmSDMA0_CNTL + sdma_offsets[i], f32_cntl);
+	}
+}
+
+/**
  * cik_sdma_enable - stop the async dma engines
  *
  * @adev: amdgpu_device pointer
@@ -382,12 +428,10 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 	struct amdgpu_ring *ring;
 	u32 rb_cntl, ib_cntl;
 	u32 rb_bufsz;
-	u32 wb_offset;
 	int i, j, r;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ring = &adev->sdma.instance[i].ring;
-		wb_offset = (ring->rptr_offs * 4);
 
 		mutex_lock(&adev->srbm_mutex);
 		for (j = 0; j < 16; j++) {
@@ -423,9 +467,9 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 
 		/* set the wb address whether it's enabled or not */
 		WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i],
-		       upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
+		       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
 		WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i],
-		       ((adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC));
+		       ((ring->rptr_gpu_addr) & 0xFFFFFFFC));
 
 		rb_cntl |= SDMA0_GFX_RB_CNTL__RPTR_WRITEBACK_ENABLE_MASK;
 
@@ -433,7 +477,7 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 		WREG32(mmSDMA0_GFX_RB_BASE_HI + sdma_offsets[i], ring->gpu_addr >> 40);
 
 		ring->wptr = 0;
-		WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], lower_32_bits(ring->wptr) << 2);
+		WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], ring->wptr << 2);
 
 		/* enable DMA RB */
 		WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i],
@@ -445,22 +489,15 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		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;
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
 			return r;
-		}
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_active_vram_size(adev, adev->mc.real_vram_size);
 	}
 
 	return 0;
@@ -537,6 +574,8 @@ static int cik_sdma_start(struct amdgpu_device *adev)
 
 	/* halt the engine before programing */
 	cik_sdma_enable(adev, false);
+	/* enable sdma ring preemption */
+	cik_ctx_switch_enable(adev, true);
 
 	/* start the gfx rings and rlc compute queues */
 	r = cik_sdma_gfx_resume(adev);
@@ -567,22 +606,18 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring)
 	u32 tmp;
 	u64 gpu_addr;
 
-	r = amdgpu_wb_get(adev, &index);
-	if (r) {
-		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
 		return r;
-	}
 
 	gpu_addr = adev->wb.gpu_addr + (index * 4);
 	tmp = 0xCAFEDEAD;
 	adev->wb.wb[index] = cpu_to_le32(tmp);
 
 	r = amdgpu_ring_alloc(ring, 5);
-	if (r) {
-		DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
-		amdgpu_wb_free(adev, index);
-		return r;
-	}
+	if (r)
+		goto error_free_wb;
+
 	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0));
 	amdgpu_ring_write(ring, lower_32_bits(gpu_addr));
 	amdgpu_ring_write(ring, upper_32_bits(gpu_addr));
@@ -594,18 +629,14 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring)
 		tmp = le32_to_cpu(adev->wb.wb[index]);
 		if (tmp == 0xDEADBEEF)
 			break;
-		DRM_UDELAY(1);
+		udelay(1);
 	}
 
-	if (i < adev->usec_timeout) {
-		DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
-	} else {
-		DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n",
-			  ring->idx, tmp);
-		r = -EINVAL;
-	}
-	amdgpu_wb_free(adev, index);
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
 
+error_free_wb:
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -613,6 +644,7 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring)
  * cik_sdma_ring_test_ib - test an IB on the DMA engine
  *
  * @ring: amdgpu_ring structure holding ring information
+ * @timeout: timeout value in jiffies, or MAX_SCHEDULE_TIMEOUT
  *
  * Test a simple IB in the DMA ring (CIK).
  * Returns 0 on success, error on failure.
@@ -627,21 +659,18 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	u64 gpu_addr;
 	long r;
 
-	r = amdgpu_wb_get(adev, &index);
-	if (r) {
-		dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
 		return r;
-	}
 
 	gpu_addr = adev->wb.gpu_addr + (index * 4);
 	tmp = 0xCAFEDEAD;
 	adev->wb.wb[index] = cpu_to_le32(tmp);
 	memset(&ib, 0, sizeof(ib));
-	r = amdgpu_ib_get(adev, NULL, 256, &ib);
-	if (r) {
-		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+	r = amdgpu_ib_get(adev, NULL, 256,
+					AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r)
 		goto err0;
-	}
 
 	ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE,
 				SDMA_WRITE_SUB_OPCODE_LINEAR, 0);
@@ -656,32 +685,27 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 
 	r = dma_fence_wait_timeout(f, false, timeout);
 	if (r == 0) {
-		DRM_ERROR("amdgpu: IB test timed out\n");
 		r = -ETIMEDOUT;
 		goto err1;
 	} else if (r < 0) {
-		DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
 		goto err1;
 	}
 	tmp = le32_to_cpu(adev->wb.wb[index]);
-	if (tmp == 0xDEADBEEF) {
-		DRM_INFO("ib test on ring %d succeeded\n", ring->idx);
+	if (tmp == 0xDEADBEEF)
 		r = 0;
-	} else {
-		DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
+	else
 		r = -EINVAL;
-	}
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
-	amdgpu_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
 /**
- * cik_sdma_vm_copy_pages - update PTEs by copying them from the GART
+ * cik_sdma_vm_copy_pte - update PTEs by copying them from the GART
  *
  * @ib: indirect buffer to fill with commands
  * @pe: addr of the page entry
@@ -707,7 +731,7 @@ static void cik_sdma_vm_copy_pte(struct amdgpu_ib *ib,
 }
 
 /**
- * cik_sdma_vm_write_pages - update PTEs by writing them manually
+ * cik_sdma_vm_write_pte - update PTEs by writing them manually
  *
  * @ib: indirect buffer to fill with commands
  * @pe: addr of the page entry
@@ -736,7 +760,7 @@ static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe,
 }
 
 /**
- * cik_sdma_vm_set_pages - update the page tables using sDMA
+ * cik_sdma_vm_set_pte_pde - update the page tables using sDMA
  *
  * @ib: indirect buffer to fill with commands
  * @pe: addr of the page entry
@@ -765,18 +789,19 @@ static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe,
 }
 
 /**
- * cik_sdma_vm_pad_ib - pad the IB to the required number of dw
+ * cik_sdma_ring_pad_ib - pad the IB to the required number of dw
  *
+ * @ring: amdgpu_ring structure holding ring information
  * @ib: indirect buffer to fill with padding
  *
  */
 static void cik_sdma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
 {
-	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
+	struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
 	u32 pad_count;
 	int i;
 
-	pad_count = (8 - (ib->length_dw & 0x7)) % 8;
+	pad_count = (-ib->length_dw) & 7;
 	for (i = 0; i < pad_count; i++)
 		if (sdma && sdma->burst_nop && (i == 0))
 			ib->ptr[ib->length_dw++] =
@@ -807,7 +832,7 @@ static void cik_sdma_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, addr & 0xfffffffc);
 	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff);
 	amdgpu_ring_write(ring, seq); /* reference */
-	amdgpu_ring_write(ring, 0xfffffff); /* mask */
+	amdgpu_ring_write(ring, 0xffffffff); /* mask */
 	amdgpu_ring_write(ring, (0xfff << 16) | 4); /* retry count, poll interval */
 }
 
@@ -815,29 +840,19 @@ static void cik_sdma_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
  * cik_sdma_ring_emit_vm_flush - cik vm flush using sDMA
  *
  * @ring: amdgpu_ring pointer
- * @vm: amdgpu_vm pointer
+ * @vmid: vmid number to use
+ * @pd_addr: address
  *
  * Update the page table base and flush the VM TLB
  * using sDMA (CIK).
  */
 static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					unsigned vm_id, uint64_t pd_addr)
+					unsigned vmid, uint64_t pd_addr)
 {
 	u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(0) |
 			  SDMA_POLL_REG_MEM_EXTRA_FUNC(0)); /* always */
 
-	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
-	if (vm_id < 8) {
-		amdgpu_ring_write(ring, (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id));
-	} else {
-		amdgpu_ring_write(ring, (mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vm_id - 8));
-	}
-	amdgpu_ring_write(ring, pd_addr >> 12);
-
-	/* flush TLB */
-	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
-	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST);
-	amdgpu_ring_write(ring, 1 << vm_id);
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
 
 	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits));
 	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST << 2);
@@ -847,6 +862,14 @@ static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, (0xfff << 16) | 10); /* retry count, poll interval */
 }
 
+static void cik_sdma_ring_emit_wreg(struct amdgpu_ring *ring,
+				    uint32_t reg, uint32_t val)
+{
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, val);
+}
+
 static void cik_enable_sdma_mgcg(struct amdgpu_device *adev,
 				 bool enable)
 {
@@ -896,12 +919,17 @@ static void cik_enable_sdma_mgls(struct amdgpu_device *adev,
 	}
 }
 
-static int cik_sdma_early_init(void *handle)
+static int cik_sdma_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
 
 	adev->sdma.num_instances = SDMA_MAX_INSTANCE;
 
+	r = cik_sdma_init_microcode(adev);
+	if (r)
+		return r;
+
 	cik_sdma_set_ring_funcs(adev);
 	cik_sdma_set_irq_funcs(adev);
 	cik_sdma_set_buffer_funcs(adev);
@@ -910,32 +938,26 @@ static int cik_sdma_early_init(void *handle)
 	return 0;
 }
 
-static int cik_sdma_sw_init(void *handle)
+static int cik_sdma_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r, i;
 
-	r = cik_sdma_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
-		return r;
-	}
-
 	/* SDMA trap event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 224,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224,
 			      &adev->sdma.trap_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 241,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 241,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 247,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 247,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
@@ -946,9 +968,9 @@ static int cik_sdma_sw_init(void *handle)
 		sprintf(ring->name, "sdma%d", i);
 		r = amdgpu_ring_init(adev, ring, 1024,
 				     &adev->sdma.trap_irq,
-				     (i == 0) ?
-				     AMDGPU_SDMA_IRQ_TRAP0 :
-				     AMDGPU_SDMA_IRQ_TRAP1);
+				     (i == 0) ? AMDGPU_SDMA_IRQ_INSTANCE0 :
+				     AMDGPU_SDMA_IRQ_INSTANCE1,
+				     AMDGPU_RING_PRIO_DEFAULT, NULL);
 		if (r)
 			return r;
 	}
@@ -956,9 +978,9 @@ static int cik_sdma_sw_init(void *handle)
 	return r;
 }
 
-static int cik_sdma_sw_fini(void *handle)
+static int cik_sdma_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
@@ -968,46 +990,38 @@ static int cik_sdma_sw_fini(void *handle)
 	return 0;
 }
 
-static int cik_sdma_hw_init(void *handle)
+static int cik_sdma_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	r = cik_sdma_start(adev);
-	if (r)
-		return r;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	return r;
+	return cik_sdma_start(adev);
 }
 
-static int cik_sdma_hw_fini(void *handle)
+static int cik_sdma_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
+	cik_ctx_switch_enable(adev, false);
 	cik_sdma_enable(adev, false);
 
 	return 0;
 }
 
-static int cik_sdma_suspend(void *handle)
+static int cik_sdma_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_sdma_hw_fini(adev);
+	return cik_sdma_hw_fini(ip_block);
 }
 
-static int cik_sdma_resume(void *handle)
+static int cik_sdma_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	cik_sdma_soft_reset(ip_block);
 
-	cik_sdma_soft_reset(handle);
-
-	return cik_sdma_hw_init(adev);
+	return cik_sdma_hw_init(ip_block);
 }
 
-static bool cik_sdma_is_idle(void *handle)
+static bool cik_sdma_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS2);
 
 	if (tmp & (SRBM_STATUS2__SDMA_BUSY_MASK |
@@ -1017,43 +1031,36 @@ static bool cik_sdma_is_idle(void *handle)
 	return true;
 }
 
-static int cik_sdma_wait_for_idle(void *handle)
+static int cik_sdma_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(mmSRBM_STATUS2) & (SRBM_STATUS2__SDMA_BUSY_MASK |
-				SRBM_STATUS2__SDMA1_BUSY_MASK);
-
-		if (!tmp)
+		if (cik_sdma_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int cik_sdma_soft_reset(void *handle)
+static int cik_sdma_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	u32 tmp = RREG32(mmSRBM_STATUS2);
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 tmp;
 
-	if (tmp & SRBM_STATUS2__SDMA_BUSY_MASK) {
-		/* sdma0 */
-		tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET);
-		tmp |= SDMA0_F32_CNTL__HALT_MASK;
-		WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp);
-		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK;
-	}
-	if (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK) {
-		/* sdma1 */
-		tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET);
-		tmp |= SDMA0_F32_CNTL__HALT_MASK;
-		WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp);
-		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK;
-	}
+	/* sdma0 */
+	tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET);
+	tmp |= SDMA0_F32_CNTL__HALT_MASK;
+	WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp);
+	srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK;
+
+	/* sdma1 */
+	tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET);
+	tmp |= SDMA0_F32_CNTL__HALT_MASK;
+	WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp);
+	srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK;
 
 	if (srbm_soft_reset) {
 		tmp = RREG32(mmSRBM_SOFT_RESET);
@@ -1083,7 +1090,7 @@ static int cik_sdma_set_trap_irq_state(struct amdgpu_device *adev,
 	u32 sdma_cntl;
 
 	switch (type) {
-	case AMDGPU_SDMA_IRQ_TRAP0:
+	case AMDGPU_SDMA_IRQ_INSTANCE0:
 		switch (state) {
 		case AMDGPU_IRQ_STATE_DISABLE:
 			sdma_cntl = RREG32(mmSDMA0_CNTL + SDMA0_REGISTER_OFFSET);
@@ -1099,7 +1106,7 @@ static int cik_sdma_set_trap_irq_state(struct amdgpu_device *adev,
 			break;
 		}
 		break;
-	case AMDGPU_SDMA_IRQ_TRAP1:
+	case AMDGPU_SDMA_IRQ_INSTANCE1:
 		switch (state) {
 		case AMDGPU_IRQ_STATE_DISABLE:
 			sdma_cntl = RREG32(mmSDMA0_CNTL + SDMA1_REGISTER_OFFSET);
@@ -1166,16 +1173,19 @@ static int cik_sdma_process_illegal_inst_irq(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *source,
 					     struct amdgpu_iv_entry *entry)
 {
+	u8 instance_id;
+
 	DRM_ERROR("Illegal instruction in SDMA command stream\n");
-	schedule_work(&adev->reset_work);
+	instance_id = (entry->ring_id & 0x3) >> 0;
+	drm_sched_fault(&adev->sdma.instance[instance_id].ring.sched);
 	return 0;
 }
 
-static int cik_sdma_set_clockgating_state(void *handle,
+static int cik_sdma_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -1186,7 +1196,7 @@ static int cik_sdma_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int cik_sdma_set_powergating_state(void *handle,
+static int cik_sdma_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1195,7 +1205,6 @@ static int cik_sdma_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cik_sdma_ip_funcs = {
 	.name = "cik_sdma",
 	.early_init = cik_sdma_early_init,
-	.late_init = NULL,
 	.sw_init = cik_sdma_sw_init,
 	.sw_fini = cik_sdma_sw_fini,
 	.hw_init = cik_sdma_hw_init,
@@ -1219,9 +1228,9 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
 	.set_wptr = cik_sdma_ring_set_wptr,
 	.emit_frame_size =
 		6 + /* cik_sdma_ring_emit_hdp_flush */
-		3 + /* cik_sdma_ring_emit_hdp_invalidate */
+		3 + /* hdp invalidate */
 		6 + /* cik_sdma_ring_emit_pipeline_sync */
-		12 + /* cik_sdma_ring_emit_vm_flush */
+		CIK_FLUSH_GPU_TLB_NUM_WREG * 3 + 6 + /* cik_sdma_ring_emit_vm_flush */
 		9 + 9 + 9, /* cik_sdma_ring_emit_fence x3 for user fence, vm fence */
 	.emit_ib_size = 7 + 4, /* cik_sdma_ring_emit_ib */
 	.emit_ib = cik_sdma_ring_emit_ib,
@@ -1229,19 +1238,21 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
 	.emit_pipeline_sync = cik_sdma_ring_emit_pipeline_sync,
 	.emit_vm_flush = cik_sdma_ring_emit_vm_flush,
 	.emit_hdp_flush = cik_sdma_ring_emit_hdp_flush,
-	.emit_hdp_invalidate = cik_sdma_ring_emit_hdp_invalidate,
 	.test_ring = cik_sdma_ring_test_ring,
 	.test_ib = cik_sdma_ring_test_ib,
 	.insert_nop = cik_sdma_ring_insert_nop,
 	.pad_ib = cik_sdma_ring_pad_ib,
+	.emit_wreg = cik_sdma_ring_emit_wreg,
 };
 
 static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	for (i = 0; i < adev->sdma.num_instances; i++)
+	for (i = 0; i < adev->sdma.num_instances; i++) {
 		adev->sdma.instance[i].ring.funcs = &cik_sdma_ring_funcs;
+		adev->sdma.instance[i].ring.me = i;
+	}
 }
 
 static const struct amdgpu_irq_src_funcs cik_sdma_trap_irq_funcs = {
@@ -1263,10 +1274,11 @@ static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev)
 /**
  * cik_sdma_emit_copy_buffer - copy buffer using the sDMA engine
  *
- * @ring: amdgpu_ring structure holding ring information
+ * @ib: indirect buffer to copy to
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
+ * @copy_flags: unused
  *
  * Copy GPU buffers using the DMA engine (CIK).
  * Used by the amdgpu ttm implementation to move pages if
@@ -1275,7 +1287,8 @@ static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev)
 static void cik_sdma_emit_copy_buffer(struct amdgpu_ib *ib,
 				      uint64_t src_offset,
 				      uint64_t dst_offset,
-				      uint32_t byte_count)
+				      uint32_t byte_count,
+				      uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0);
 	ib->ptr[ib->length_dw++] = byte_count;
@@ -1289,7 +1302,7 @@ static void cik_sdma_emit_copy_buffer(struct amdgpu_ib *ib,
 /**
  * cik_sdma_emit_fill_buffer - fill buffer using the sDMA engine
  *
- * @ring: amdgpu_ring structure holding ring information
+ * @ib: indirect buffer to fill
  * @src_data: value to write to buffer
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
@@ -1320,14 +1333,14 @@ static const struct amdgpu_buffer_funcs cik_sdma_buffer_funcs = {
 
 static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mman.buffer_funcs == NULL) {
-		adev->mman.buffer_funcs = &cik_sdma_buffer_funcs;
-		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
-	}
+	adev->mman.buffer_funcs = &cik_sdma_buffer_funcs;
+	adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
 }
 
 static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = {
+	.copy_pte_num_dw = 7,
 	.copy_pte = cik_sdma_vm_copy_pte,
+
 	.write_pte = cik_sdma_vm_write_pte,
 	.set_pte_pde = cik_sdma_vm_set_pte_pde,
 };
@@ -1336,14 +1349,12 @@ static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev)
 {
 	unsigned i;
 
-	if (adev->vm_manager.vm_pte_funcs == NULL) {
-		adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++)
-			adev->vm_manager.vm_pte_rings[i] =
-				&adev->sdma.instance[i].ring;
-
-		adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances;
+	adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		adev->vm_manager.vm_pte_scheds[i] =
+			&adev->sdma.instance[i].ring.sched;
 	}
+	adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances;
 }
 
 const struct amdgpu_ip_block_version cik_sdma_ip_block =