[PATCH 6/6] drm/amdgpu: Implement ib_preemmpt interface for sdma

[Rex Zhu]

sdma can be preempted via this interface

Signed-off-by: Rex Zhu <Rex.Zhu at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 14 ++++++++++++++
 drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 15 +++++++++++++++
 2 files changed, 29 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
index de280d4..388d3eb 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
@@ -545,6 +545,19 @@ static void sdma_v3_0_ring_patch_cond_exec(struct amdgpu_ring *ring, unsigned of
 		ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
 }
 
+static void sdma_v3_0_ring_ib_preempt(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 index = 0;
+
+	if (!amdgpu_sriov_vf(adev) && adev->gpu_preemption) {
+		amdgpu_ring_alloc(ring, 16);
+		amdgpu_ring_set_preempt_cond_exec(ring, true);
+		WREG32(index == 0 ? mmSDMA0_GFX_PREEMPT : mmSDMA1_GFX_PREEMPT, 0x1);
+		amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr+8, 0xff, AMDGPU_FENCE_FLAG_CLEAR_PREEMPT);
+		amdgpu_ring_commit(ring);
+	}
+}
 
 /**
  * sdma_v3_0_gfx_stop - stop the gfx async dma engines
@@ -1660,6 +1673,7 @@ static void sdma_v3_0_get_clockgating_state(void *handle, u32 *flags)
 	.emit_wreg = sdma_v3_0_ring_emit_wreg,
 	.init_cond_exec = sdma_v3_0_ring_init_cond_exec,
 	.patch_cond_exec = sdma_v3_0_ring_patch_cond_exec,
+	.ib_preempt = sdma_v3_0_ring_ib_preempt,
 };
 
 static void sdma_v3_0_set_ring_funcs(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 33bdeeb..f47a3d9 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -654,6 +654,20 @@ static void sdma_v4_0_ring_patch_cond_exec(struct amdgpu_ring *ring, unsigned of
 		ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
 }
 
+static void sdma_v4_0_ring_ib_preempt(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 index = 0;
+
+	if (!amdgpu_sriov_vf(adev) && adev->gpu_preemption) {
+		amdgpu_ring_alloc(ring, 16);
+		amdgpu_ring_set_preempt_cond_exec(ring, true);
+		WREG32(index == 0 ? mmSDMA0_GFX_PREEMPT : mmSDMA1_GFX_PREEMPT, 0x1);
+		amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr+8, 0xff, AMDGPU_FENCE_FLAG_CLEAR_PREEMPT);
+		amdgpu_ring_commit(ring);
+	}
+}
+
 /**
  * sdma_v4_0_gfx_stop - stop the gfx async dma engines
  *
@@ -1996,6 +2010,7 @@ static void sdma_v4_0_get_clockgating_state(void *handle, u32 *flags)
 	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
 	.init_cond_exec = sdma_v4_0_ring_init_cond_exec,
 	.patch_cond_exec = sdma_v4_0_ring_patch_cond_exec,
+	.ib_preempt = sdma_v4_0_ring_ib_preempt,
 };
 
 static const struct amdgpu_ring_funcs sdma_v4_0_page_ring_funcs = {
-- 
1.9.1