[PATCH 15/15] drm/amdgpu: enable BACO reset for SMU7 based dGPUs (v2)

[Alex Deucher]

Use BACO to reset the GPU if supported on SMU7 based
dGPUs.

v2: don't use baco on CI parts

Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/cik.c | 48 ++++++++++++++++--
 drivers/gpu/drm/amd/amdgpu/cik.h |  3 ++
 drivers/gpu/drm/amd/amdgpu/vi.c  | 84 ++++++++++++++++++++++++++++++--
 drivers/gpu/drm/amd/amdgpu/vi.h  |  3 ++
 4 files changed, 128 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c
index b81bb414fcb3..fc8b34480f66 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik.c
@@ -1270,15 +1270,15 @@ static int cik_gpu_pci_config_reset(struct amdgpu_device *adev)
 }
 
 /**
- * cik_asic_reset - soft reset GPU
+ * cik_asic_pci_config_reset - soft reset GPU
  *
  * @adev: amdgpu_device pointer
  *
- * Look up which blocks are hung and attempt
- * to reset them.
+ * Use PCI Config method to reset the GPU.
+ *
  * Returns 0 for success.
  */
-static int cik_asic_reset(struct amdgpu_device *adev)
+static int cik_asic_pci_config_reset(struct amdgpu_device *adev)
 {
 	int r;
 
@@ -1294,7 +1294,45 @@ static int cik_asic_reset(struct amdgpu_device *adev)
 static enum amd_reset_method
 cik_asic_reset_method(struct amdgpu_device *adev)
 {
-	return AMD_RESET_METHOD_LEGACY;
+	bool baco_reset;
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+	case CHIP_HAWAII:
+		/* disable baco reset until it works */
+		/* smu7_asic_get_baco_capability(adev, &baco_reset); */
+		baco_reset = false;
+		break;
+	default:
+		baco_reset = false;
+		break;
+	}
+
+	if (baco_reset)
+		return AMD_RESET_METHOD_BACO;
+	else
+		return AMD_RESET_METHOD_LEGACY;
+}
+
+/**
+ * cik_asic_reset - soft reset GPU
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Look up which blocks are hung and attempt
+ * to reset them.
+ * Returns 0 for success.
+ */
+static int cik_asic_reset(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (cik_asic_reset_method(adev) == AMD_RESET_METHOD_BACO)
+		r = smu7_asic_baco_reset(adev);
+	else
+		r = cik_asic_pci_config_reset(adev);
+
+	return r;
 }
 
 static u32 cik_get_config_memsize(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/cik.h b/drivers/gpu/drm/amd/amdgpu/cik.h
index 54c625a2e570..9870bf27870e 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik.h
+++ b/drivers/gpu/drm/amd/amdgpu/cik.h
@@ -31,4 +31,7 @@ void cik_srbm_select(struct amdgpu_device *adev,
 int cik_set_ip_blocks(struct amdgpu_device *adev);
 
 void legacy_doorbell_index_init(struct amdgpu_device *adev);
+int smu7_asic_get_baco_capability(struct amdgpu_device *adev, bool *cap);
+int smu7_asic_baco_reset(struct amdgpu_device *adev);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c
index 5f8c8786cac5..78e5cdc0c058 100644
--- a/drivers/gpu/drm/amd/amdgpu/vi.c
+++ b/drivers/gpu/drm/amd/amdgpu/vi.c
@@ -689,16 +689,50 @@ static int vi_gpu_pci_config_reset(struct amdgpu_device *adev)
 	return -EINVAL;
 }
 
+int smu7_asic_get_baco_capability(struct amdgpu_device *adev, bool *cap)
+{
+	void *pp_handle = adev->powerplay.pp_handle;
+	const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs;
+
+	if (!pp_funcs || !pp_funcs->get_asic_baco_capability) {
+		*cap = false;
+		return -ENOENT;
+	}
+
+	return pp_funcs->get_asic_baco_capability(pp_handle, cap);
+}
+
+int smu7_asic_baco_reset(struct amdgpu_device *adev)
+{
+	void *pp_handle = adev->powerplay.pp_handle;
+	const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs;
+
+	if (!pp_funcs ||!pp_funcs->get_asic_baco_state ||!pp_funcs->set_asic_baco_state)
+		return -ENOENT;
+
+	/* enter BACO state */
+	if (pp_funcs->set_asic_baco_state(pp_handle, 1))
+		return -EIO;
+
+	/* exit BACO state */
+	if (pp_funcs->set_asic_baco_state(pp_handle, 0))
+		return -EIO;
+
+	dev_info(adev->dev, "GPU BACO reset\n");
+
+	return 0;
+}
+
 /**
- * vi_asic_reset - soft reset GPU
+ * vi_asic_pci_config_reset - soft reset GPU
  *
  * @adev: amdgpu_device pointer
  *
- * Look up which blocks are hung and attempt
- * to reset them.
+ * Use PCI Config method to reset the GPU.
+ *
  * Returns 0 for success.
  */
-static int vi_asic_reset(struct amdgpu_device *adev)
+static int vi_asic_pci_config_reset(struct amdgpu_device *adev)
 {
 	int r;
 
@@ -714,7 +748,47 @@ static int vi_asic_reset(struct amdgpu_device *adev)
 static enum amd_reset_method
 vi_asic_reset_method(struct amdgpu_device *adev)
 {
-	return AMD_RESET_METHOD_LEGACY;
+	bool baco_reset;
+
+	switch (adev->asic_type) {
+	case CHIP_FIJI:
+	case CHIP_TONGA:
+	case CHIP_POLARIS10:
+	case CHIP_POLARIS11:
+	case CHIP_POLARIS12:
+	case CHIP_TOPAZ:
+		smu7_asic_get_baco_capability(adev, &baco_reset);
+		break;
+	default:
+		baco_reset = false;
+		break;
+	}
+
+	if (baco_reset)
+		return AMD_RESET_METHOD_BACO;
+	else
+		return AMD_RESET_METHOD_LEGACY;
+}
+
+/**
+ * vi_asic_reset - soft reset GPU
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Look up which blocks are hung and attempt
+ * to reset them.
+ * Returns 0 for success.
+ */
+static int vi_asic_reset(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (vi_asic_reset_method(adev) == AMD_RESET_METHOD_BACO)
+		r = smu7_asic_baco_reset(adev);
+	else
+		r = vi_asic_pci_config_reset(adev);
+
+	return r;
 }
 
 static u32 vi_get_config_memsize(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/vi.h b/drivers/gpu/drm/amd/amdgpu/vi.h
index 8de0772f986c..40d4174913a4 100644
--- a/drivers/gpu/drm/amd/amdgpu/vi.h
+++ b/drivers/gpu/drm/amd/amdgpu/vi.h
@@ -31,4 +31,7 @@ void vi_srbm_select(struct amdgpu_device *adev,
 int vi_set_ip_blocks(struct amdgpu_device *adev);
 
 void legacy_doorbell_index_init(struct amdgpu_device *adev);
+int smu7_asic_get_baco_capability(struct amdgpu_device *adev, bool *cap);
+int smu7_asic_baco_reset(struct amdgpu_device *adev);
+
 #endif
-- 
2.20.1