SWDEV-444567 - Added Ring Hang Event

Signed-off-by: Maisam Arif <maisarif@amd.com> Change-Id: I2e73ba08ee0004f6f30660b2fa425ea94bafceca [ROCm/amdsmi commit: 52843152a5]
2024-05-02 01:48:14 -05:00
@@ -8,6 +8,9 @@ Full documentation for amd_smi_lib is available at [https://rocm.docs.amd.com/](

 ### Added

+- **Added Ring Hang event**  
+Added `AMDSMI_EVT_NOTIF_RING_HANG` to the possible events in the `amdsmi_evt_notification_type_t` enum.
+
 - **Added process isolation and clean shader APIs and CLI commands**  
 Added APIs CLI and APIs to address LeftoverLocals security issues. Allowing clearing the sram data and setting process isolation on a per GPU basis. New APIs:
  - `amdsmi_get_gpu_process_isolation()`
@@ -136,6 +139,10 @@ Updates required `amdsmi_get_power_cap_info` to return in uW as originally refle
 - **Fixed python interface call amdsmi_get_gpu_memory_reserved_pages & amdsmi_get_gpu_bad_page_info**  
 Previously python interface calls to populated bad pages resulted in a `ValueError: NULL pointer access`. This fixes the bad-pages subcommand CLI  subcommand as well.

+### Known Issues
+
+- N/A
+
 ## amd_smi_lib for ROCm 6.1.1

 ### Added
@@ -835,13 +835,15 @@ typedef struct {
 * Event notification event types
 */
 typedef enum {
+  AMDSMI_EVT_NOTIF_NONE = 0,     //!< Not used
  AMDSMI_EVT_NOTIF_VMFAULT = 1,  //!< VM page fault
  AMDSMI_EVT_NOTIF_FIRST = AMDSMI_EVT_NOTIF_VMFAULT,
  AMDSMI_EVT_NOTIF_THERMAL_THROTTLE = 2,
  AMDSMI_EVT_NOTIF_GPU_PRE_RESET = 3,
  AMDSMI_EVT_NOTIF_GPU_POST_RESET = 4,
+  AMDSMI_EVT_NOTIF_RING_HANG = 5,

-  AMDSMI_EVT_NOTIF_LAST = AMDSMI_EVT_NOTIF_GPU_POST_RESET
+  AMDSMI_EVT_NOTIF_LAST = AMDSMI_EVT_NOTIF_RING_HANG
 } amdsmi_evt_notification_type_t;

 /**
@@ -1145,8 +1145,9 @@ Event Type | Description
 ---|------
 `VMFAULT` | VM page fault
 `THERMAL_THROTTLE` | thermal throttle
-`GPU_PRE_RESET`   | gpu pre reset
+`GPU_PRE_RESET` | gpu pre reset
 `GPU_POST_RESET` | gpu post reset
+`RING_HANG` | ring hang event

 #### read

@@ -217,10 +217,12 @@ class AmdSmiCounterCommand(IntEnum):


 class AmdSmiEvtNotificationType(IntEnum):
+    NONE = amdsmi_wrapper.AMDSMI_EVT_NOTIF_NONE
    VMFAULT = amdsmi_wrapper.AMDSMI_EVT_NOTIF_VMFAULT
    THERMAL_THROTTLE = amdsmi_wrapper.AMDSMI_EVT_NOTIF_THERMAL_THROTTLE
    GPU_PRE_RESET = amdsmi_wrapper.AMDSMI_EVT_NOTIF_GPU_PRE_RESET
    GPU_POST_RESET = amdsmi_wrapper.AMDSMI_EVT_NOTIF_GPU_POST_RESET
+    RING_HANG = amdsmi_wrapper.AMDSMI_EVT_NOTIF_RING_HANG


 class AmdSmiTemperatureMetric(IntEnum):
@@ -1157,19 +1157,23 @@ amdsmi_counter_value_t = struct_amdsmi_counter_value_t

 # values for enumeration 'amdsmi_evt_notification_type_t'
 amdsmi_evt_notification_type_t__enumvalues = {
+    0: 'AMDSMI_EVT_NOTIF_NONE',
    1: 'AMDSMI_EVT_NOTIF_VMFAULT',
    1: 'AMDSMI_EVT_NOTIF_FIRST',
    2: 'AMDSMI_EVT_NOTIF_THERMAL_THROTTLE',
    3: 'AMDSMI_EVT_NOTIF_GPU_PRE_RESET',
    4: 'AMDSMI_EVT_NOTIF_GPU_POST_RESET',
-    4: 'AMDSMI_EVT_NOTIF_LAST',
+    5: 'AMDSMI_EVT_NOTIF_RING_HANG',
+    5: 'AMDSMI_EVT_NOTIF_LAST',
 }
+AMDSMI_EVT_NOTIF_NONE = 0
 AMDSMI_EVT_NOTIF_VMFAULT = 1
 AMDSMI_EVT_NOTIF_FIRST = 1
 AMDSMI_EVT_NOTIF_THERMAL_THROTTLE = 2
 AMDSMI_EVT_NOTIF_GPU_PRE_RESET = 3
 AMDSMI_EVT_NOTIF_GPU_POST_RESET = 4
-AMDSMI_EVT_NOTIF_LAST = 4
+AMDSMI_EVT_NOTIF_RING_HANG = 5
+AMDSMI_EVT_NOTIF_LAST = 5
 amdsmi_evt_notification_type_t = ctypes.c_uint32 # enum
 class struct_amdsmi_evt_notification_data_t(Structure):
    pass
@@ -2393,6 +2397,7 @@ __all__ = \
    'AMDSMI_EVNT_XGMI_LAST', 'AMDSMI_EVT_NOTIF_FIRST',
    'AMDSMI_EVT_NOTIF_GPU_POST_RESET',
    'AMDSMI_EVT_NOTIF_GPU_PRE_RESET', 'AMDSMI_EVT_NOTIF_LAST',
+    'AMDSMI_EVT_NOTIF_NONE', 'AMDSMI_EVT_NOTIF_RING_HANG',
    'AMDSMI_EVT_NOTIF_THERMAL_THROTTLE', 'AMDSMI_EVT_NOTIF_VMFAULT',
    'AMDSMI_FILE_NOT_FOUND', 'AMDSMI_FREQ_IND_INVALID',
    'AMDSMI_FREQ_IND_MAX', 'AMDSMI_FREQ_IND_MIN',
@@ -553,6 +553,7 @@ enum kfd_smi_event {
  KFD_SMI_EVENT_THERMAL_THROTTLE = 2,
  KFD_SMI_EVENT_GPU_PRE_RESET = 3,
  KFD_SMI_EVENT_GPU_POST_RESET = 4,
+  KFD_SMI_EVENT_RING_HANG = 5,
 };

 #define KFD_SMI_EVENT_MASK_FROM_INDEX(i) (1ULL << ((i) - 1))
@@ -357,13 +357,16 @@ typedef struct {
 * Event notification event types
 */
 typedef enum {
+  RSMI_EVT_NOTIF_NONE = KFD_SMI_EVENT_NONE,        //!< Unused
  RSMI_EVT_NOTIF_VMFAULT = KFD_SMI_EVENT_VMFAULT,  //!< VM page fault
  RSMI_EVT_NOTIF_FIRST = RSMI_EVT_NOTIF_VMFAULT,
+
  RSMI_EVT_NOTIF_THERMAL_THROTTLE = KFD_SMI_EVENT_THERMAL_THROTTLE,
  RSMI_EVT_NOTIF_GPU_PRE_RESET = KFD_SMI_EVENT_GPU_PRE_RESET,
  RSMI_EVT_NOTIF_GPU_POST_RESET = KFD_SMI_EVENT_GPU_POST_RESET,
+  RSMI_EVT_NOTIF_RING_HANG = KFD_SMI_EVENT_RING_HANG,

-  RSMI_EVT_NOTIF_LAST = RSMI_EVT_NOTIF_GPU_POST_RESET
+  RSMI_EVT_NOTIF_LAST = RSMI_EVT_NOTIF_RING_HANG
 } rsmi_evt_notification_type_t;

 /**
@@ -134,16 +134,18 @@ class rsmi_dev_perf_level_t(c_int):
    RSMI_DEV_PERF_LEVEL_UNKNOWN = 0x100


-notification_type_names = ['VM_FAULT', 'THERMAL_THROTTLE', 'GPU_RESET']
+notification_type_names = ['VM_FAULT', 'THERMAL_THROTTLE', 'GPU_PRE_RESET', 'GPU_POST_RESET', 'RING_HANG']


 class rsmi_evt_notification_type_t(c_int):
-    RSMI_EVT_NOTIF_VMFAULT = 0
+    RSMI_EVT_NOTIF_NONE = 0
+    RSMI_EVT_NOTIF_VMFAULT = 1
    RSMI_EVT_NOTIF_FIRST = RSMI_EVT_NOTIF_VMFAULT
-    RSMI_EVT_NOTIF_THERMAL_THROTTLE = 1
-    RSMI_EVT_NOTIF_GPU_PRE_RESET = 2
-    RSMI_EVT_NOTIF_GPU_POST_RESET = 3
-    RSMI_EVT_NOTIF_LAST = RSMI_EVT_NOTIF_GPU_POST_RESET
+    RSMI_EVT_NOTIF_THERMAL_THROTTLE = 2
+    RSMI_EVT_NOTIF_GPU_PRE_RESET = 3
+    RSMI_EVT_NOTIF_GPU_POST_RESET = 4
+    RSMI_EVT_NOTIF_RING_HANG = 5
+    RSMI_EVT_NOTIF_LAST = RSMI_EVT_NOTIF_RING_HANG


 class rsmi_voltage_metric_t(c_int):
@@ -84,6 +84,7 @@ static const std::map<amdsmi_evt_notification_type_t, const char *>
    {AMDSMI_EVT_NOTIF_THERMAL_THROTTLE, "AMDSMI_EVT_NOTIF_THERMAL_THROTTLE"},
    {AMDSMI_EVT_NOTIF_GPU_PRE_RESET, "AMDSMI_EVT_NOTIF_GPU_PRE_RESET"},
    {AMDSMI_EVT_NOTIF_GPU_POST_RESET, "AMDSMI_EVT_NOTIF_GPU_POST_RESET"},
+    {AMDSMI_EVT_NOTIF_RING_HANG, "AMDSMI_EVT_NOTIF_RING_HANG"},
 };
 const char *
 NameFromEvtNotifType(amdsmi_evt_notification_type_t evt) {