drm/amdgpu: disable ras query and iject during gpu reset

From John Clements
50a8cfb0e71644f2882dbfaf93150184094245bd in linux 5.7.y/5.7.8
61380faa4b4cc577df8a7ff5db5859bac6b351f7 in mainline linux

1 files changed, 21 insertions, 3 deletions

diff --git a/sys/dev/pci/drm/amd/amdgpu/amdgpu_ras.c b/sys/dev/pci/drm/amd/amdgpu/amdgpu_ras.c
index a0e4b39add3..864fab7b049 100644
--- a/sys/dev/pci/drm/amd/amdgpu/amdgpu_ras.c
+++ b/sys/dev/pci/drm/amd/amdgpu/amdgpu_ras.c
@@ -80,6 +80,20 @@ atomic_t amdgpu_ras_in_intr = ATOMIC_INIT(0);
 static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
 				uint64_t addr);
 
+void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready)
+{
+	if (adev)
+		amdgpu_ras_get_context(adev)->error_query_ready = ready;
+}
+
+bool amdgpu_ras_get_error_query_ready(struct amdgpu_device *adev)
+{
+	if (adev)
+		return amdgpu_ras_get_context(adev)->error_query_ready;
+
+	return false;
+}
+
 #ifdef __linux__
 
 static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
@@ -283,7 +297,7 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, const char __user *
 	struct ras_debug_if data;
 	int ret = 0;
 
-	if (amdgpu_ras_intr_triggered()) {
+	if (!amdgpu_ras_get_error_query_ready(adev)) {
 		DRM_WARN("RAS WARN: error injection currently inaccessible\n");
 		return size;
 	}
@@ -401,7 +415,7 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev,
 		.head = obj->head,
 	};
 
-	if (amdgpu_ras_intr_triggered())
+	if (!amdgpu_ras_get_error_query_ready(obj->adev))
 		return snprintf(buf, PAGE_SIZE,
 				"Query currently inaccessible\n");
 
@@ -1915,8 +1929,10 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev,
 	}
 
 	/* in resume phase, no need to create ras fs node */
-	if (adev->in_suspend || adev->in_gpu_reset)
+	if (adev->in_suspend || adev->in_gpu_reset) {
+		amdgpu_ras_set_error_query_ready(adev, true);
 		return 0;
+	}
 
 	if (ih_info->cb) {
 		r = amdgpu_ras_interrupt_add_handler(adev, ih_info);
@@ -1928,6 +1944,8 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev,
 	if (r)
 		goto sysfs;
 
+	amdgpu_ras_set_error_query_ready(adev, true);
+
 	return 0;
 cleanup:
 	amdgpu_ras_sysfs_remove(adev, ras_block);