drm/amdgpu: add work function for GPU reset event

This patch adds a work function, which sends a GPU reset uevent and some contextual infomration, like the PID and some status flags. This work should be scheduled during a GPU reset. The userspace can do some recovery and post-processing work based on this event and information. V2: Addressed review comments from Christian - Changed the name of the work to gpu_reset_event_work - Added a structure to accommodate some additional information (like a PID and some flags) - Do not add new structure in amdgpu.h Cc: Alexander Deucher <[email protected]> Cc: Christian Koenig <[email protected]> Cc: Amaranath Somalapuram <[email protected]> Signed-off-by: Shashank Sharma <[email protected]> (cherry picked from commit f63b09e) [Forward port to 6.0] Signed-off-by: Cristian Ciocaltea <[email protected]>
ChimeraOS · alkazar · Sep 16, 2019 · May 19, 2022 · Jan 13, 2024 · Dec 12, 2023
commit 945b74ba992c8c0fbca2a1c45a0eeb665a29ee59
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -57,6 +57,7 @@
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_gem.h>
 #include <drm/drm_ioctl.h>
+#include <drm/drm_sysfs.h>
 
 #include <kgd_kfd_interface.h>
 #include "dm_pp_interface.h"
@@ -1101,6 +1102,7 @@ struct amdgpu_device {
 
 	int asic_reset_res;
 	struct work_struct		xgmi_reset_work;
+	struct work_struct		gpu_reset_event_work;
 	struct list_head		reset_list;
 
 	long				gfx_timeout;
@@ -1132,6 +1134,7 @@ struct amdgpu_device {
 	bool				barrier_has_auto_waitcnt;
 
 	struct amdgpu_reset_control     *reset_cntl;
+	struct drm_reset_event		reset_event_info;
 	uint32_t                        ip_versions[MAX_HWIP][HWIP_MAX_INSTANCE];
 
 	bool				ram_is_direct_mapped;

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -81,6 +81,7 @@
 #include <linux/pm_runtime.h>
 
 #include <drm/drm_drv.h>
+#include <drm/drm_sysfs.h>
 
 #if IS_ENABLED(CONFIG_X86)
 #include <asm/intel-family.h>
@@ -3816,6 +3817,17 @@ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev)
 	return amdgpu_device_asic_has_dc_support(adev->asic_type);
 }
 
+static void amdgpu_device_reset_event_func(struct work_struct *__work)
+{
+	struct amdgpu_device *adev = container_of(__work, struct amdgpu_device,
+						  gpu_reset_event_work);
+	/*
+	 * A GPU reset has happened, inform the userspace and pass the
+	 * reset related information.
+	 */
+	drm_sysfs_reset_event(&adev->ddev, &adev->reset_event_info);
+}
+
 static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
 {
 	struct amdgpu_device *adev =
@@ -4086,6 +4098,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 			  amdgpu_device_delay_enable_gfx_off);
 
 	INIT_WORK(&adev->xgmi_reset_work, amdgpu_device_xgmi_reset_func);
+	INIT_WORK(&adev->gpu_reset_event_work, amdgpu_device_reset_event_func);
 
 	adev->gfx.gfx_off_req_count = 1;
 	adev->gfx.gfx_off_residency = 0;