[PATCH v2] drm/amd/amdgpu/vcn_v2_0: Mark RB commands as KMD commands

[Thai, Thong]

Sets the CMD_SOURCE bit for VCN 2.0 Decoder Ring Buffer commands. This
bit was previously set by the RBC HW on older firmware. Newer firmware
uses a SW RBC and this bit has to be set by the driver.

Signed-off-by: Thong Thai <thong.thai at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h |  1 +
 drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c   | 44 +++++++++++++++++++++----
 2 files changed, 38 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
index 5e2453ee6b29..4d3bf4adf1eb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
@@ -30,6 +30,7 @@
 #define AMDGPU_VCN_FIRMWARE_OFFSET	256
 #define AMDGPU_VCN_MAX_ENC_RINGS	3
 
+#define VCN_DEC_KMD_CMD			0x80000000
 #define VCN_DEC_CMD_FENCE		0x00000000
 #define VCN_DEC_CMD_TRAP		0x00000001
 #define VCN_DEC_CMD_WRITE_REG		0x00000004
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
index bc9726787c97..de928f1b1528 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
@@ -1488,7 +1488,7 @@ static void vcn_v2_0_dec_ring_insert_start(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET, 0));
 	amdgpu_ring_write(ring, 0);
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET, 0));
-	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_START << 1);
+	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_PACKET_START << 1));
 }
 
 /**
@@ -1501,7 +1501,7 @@ static void vcn_v2_0_dec_ring_insert_start(struct amdgpu_ring *ring)
 static void vcn_v2_0_dec_ring_insert_end(struct amdgpu_ring *ring)
 {
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET, 0));
-	amdgpu_ring_write(ring, VCN_DEC_CMD_PACKET_END << 1);
+	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_PACKET_END << 1));
 }
 
 /**
@@ -1546,7 +1546,7 @@ static void vcn_v2_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
 	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff);
 
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET, 0));
-	amdgpu_ring_write(ring, VCN_DEC_CMD_FENCE << 1);
+	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_FENCE << 1));
 
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0_INTERNAL_OFFSET, 0));
 	amdgpu_ring_write(ring, 0);
@@ -1556,7 +1556,7 @@ static void vcn_v2_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
 
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET, 0));
 
-	amdgpu_ring_write(ring, VCN_DEC_CMD_TRAP << 1);
+	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_TRAP << 1));
 }
 
 /**
@@ -1600,7 +1600,7 @@ static void vcn_v2_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring,
 
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET, 0));
 
-	amdgpu_ring_write(ring, VCN_DEC_CMD_REG_READ_COND_WAIT << 1);
+	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_REG_READ_COND_WAIT << 1));
 }
 
 static void vcn_v2_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
@@ -1629,7 +1629,7 @@ static void vcn_v2_0_dec_ring_emit_wreg(struct amdgpu_ring *ring,
 
 	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD_INTERNAL_OFFSET, 0));
 
-	amdgpu_ring_write(ring, VCN_DEC_CMD_WRITE_REG << 1);
+	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_WRITE_REG << 1));
 }
 
 /**
@@ -2082,6 +2082,36 @@ static int vcn_v2_0_process_interrupt(struct amdgpu_device *adev,
 	return 0;
 }
 
+int vcn_v2_0_dec_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	WREG32(adev->vcn.external.scratch9, 0xCAFEDEAD);
+	r = amdgpu_ring_alloc(ring, 4);
+	if (r)
+		return r;
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.cmd, 0));
+	amdgpu_ring_write(ring, VCN_DEC_KMD_CMD | (VCN_DEC_CMD_PACKET_START << 1));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.scratch9, 0));
+	amdgpu_ring_write(ring, 0xDEADBEEF);
+	amdgpu_ring_commit(ring);
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(adev->vcn.external.scratch9);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+
+	return r;
+}
+
+
 static int vcn_v2_0_set_powergating_state(void *handle,
 					  enum amd_powergating_state state)
 {
@@ -2145,7 +2175,7 @@ static const struct amdgpu_ring_funcs vcn_v2_0_dec_ring_vm_funcs = {
 	.emit_ib = vcn_v2_0_dec_ring_emit_ib,
 	.emit_fence = vcn_v2_0_dec_ring_emit_fence,
 	.emit_vm_flush = vcn_v2_0_dec_ring_emit_vm_flush,
-	.test_ring = amdgpu_vcn_dec_ring_test_ring,
+	.test_ring = vcn_v2_0_dec_ring_test_ring,
 	.test_ib = amdgpu_vcn_dec_ring_test_ib,
 	.insert_nop = vcn_v2_0_dec_ring_insert_nop,
 	.insert_start = vcn_v2_0_dec_ring_insert_start,
-- 
2.17.1