1 files changed, 34 insertions, 17 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/aldebaran.c b/drivers/gpu/drm/amd/amdgpu/aldebaran.c
index bcfdb63b1d42..2b97b8a96fb4 100644
--- a/drivers/gpu/drm/amd/amdgpu/aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/aldebaran.c
@@ -31,6 +31,17 @@
 #include "amdgpu_psp.h"
 #include "amdgpu_xgmi.h"
 
+static bool aldebaran_is_mode2_default(struct amdgpu_reset_control *reset_ctl)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+
+	if ((adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	     adev->gmc.xgmi.connected_to_cpu))
+		return true;
+
+	return false;
+}
+
 static struct amdgpu_reset_handler *
 aldebaran_get_reset_handler(struct amdgpu_reset_control *reset_ctl,
 			    struct amdgpu_reset_context *reset_context)
@@ -48,7 +59,7 @@ aldebaran_get_reset_handler(struct amdgpu_reset_control *reset_ctl,
 		}
 	}
 
-	if (adev->gmc.xgmi.connected_to_cpu) {
+	if (aldebaran_is_mode2_default(reset_ctl)) {
 		list_for_each_entry(handler, &reset_ctl->reset_handlers,
 				     handler_list) {
 			if (handler->reset_method == AMD_RESET_METHOD_MODE2) {
@@ -136,18 +147,23 @@ static int
 aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 			      struct amdgpu_reset_context *reset_context)
 {
-	struct amdgpu_device *tmp_adev = NULL;
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *tmp_adev = NULL;
 	int r = 0;
 
 	dev_dbg(adev->dev, "aldebaran perform hw reset\n");
-	if (reset_context->hive == NULL) {
+
+	if (reset_device_list == NULL)
+		return -EINVAL;
+
+	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
 		return -EINVAL;
 	}
 
-	list_for_each_entry(tmp_adev, &reset_context->hive->device_list,
-			     gmc.xgmi.head) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		mutex_lock(&tmp_adev->reset_cntl->reset_lock);
 		tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_MODE2;
 	}
@@ -155,8 +171,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 	 * Mode2 reset doesn't need any sync between nodes in XGMI hive, instead launch
 	 * them together so that they can be completed asynchronously on multiple nodes
 	 */
-	list_for_each_entry(tmp_adev, &reset_context->hive->device_list,
-			     gmc.xgmi.head) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		/* For XGMI run all resets in parallel to speed up the process */
 		if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
 			if (!queue_work(system_unbound_wq,
@@ -174,9 +189,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 
 	/* For XGMI wait for all resets to complete before proceed */
 	if (!r) {
-		list_for_each_entry(tmp_adev,
-				     &reset_context->hive->device_list,
-				     gmc.xgmi.head) {
+		list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 			if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
 				flush_work(&tmp_adev->reset_cntl->reset_work);
 				r = tmp_adev->asic_reset_res;
@@ -186,8 +199,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 		}
 	}
 
-	list_for_each_entry(tmp_adev, &reset_context->hive->device_list,
-			     gmc.xgmi.head) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		mutex_unlock(&tmp_adev->reset_cntl->reset_lock);
 		tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_NONE;
 	}
@@ -260,7 +272,7 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 	adev->gfx.rlc.funcs->resume(adev);
 
 	/* Wait for FW reset event complete */
-	r = smu_wait_for_event(adev, SMU_EVENT_RESET_COMPLETE, 0);
+	r = amdgpu_dpm_wait_for_event(adev, SMU_EVENT_RESET_COMPLETE, 0);
 	if (r) {
 		dev_err(adev->dev,
 			"Failed to get response from firmware after reset\n");
@@ -319,16 +331,21 @@ static int
 aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 				  struct amdgpu_reset_context *reset_context)
 {
-	int r;
+	struct list_head *reset_device_list = reset_context->reset_device_list;
 	struct amdgpu_device *tmp_adev = NULL;
+	int r;
+
+	if (reset_device_list == NULL)
+		return -EINVAL;
 
-	if (reset_context->hive == NULL) {
+	if (reset_context->reset_req_dev->ip_versions[MP1_HWIP][0] ==
+		    IP_VERSION(13, 0, 2) &&
+	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
 		return -EINVAL;
 	}
 
-	list_for_each_entry(tmp_adev, &reset_context->hive->device_list,
-			     gmc.xgmi.head) {
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
 		dev_info(tmp_adev->dev,
 			 "GPU reset succeeded, trying to resume\n");
 		r = aldebaran_mode2_restore_ip(tmp_adev);