support HMM profiling event

include/rocm_smi/kfd_ioctl.h

@@ -547,9 +547,47 @@ 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_MIGRATE_START = 5,
+  KFD_SMI_EVENT_MIGRATE_END = 6,
+  KFD_SMI_EVENT_PAGE_FAULT_START = 7,
+  KFD_SMI_EVENT_PAGE_FAULT_END = 8,
+  KFD_SMI_EVENT_QUEUE_EVICTION = 9,
+  KFD_SMI_EVENT_QUEUE_RESTORE = 10,
+  KFD_SMI_EVENT_UNMAP_FROM_GPU = 11,
+
+  /*
+   * max event number, as a flag bit to get events from all processes,
+   * this requires super user permission, otherwise will not be able to
+   * receive event from any process. Without this flag to receive events
+   * from same process.
+   */
+  KFD_SMI_EVENT_ALL_PROCESS = 64
 };
 
 #define KFD_SMI_EVENT_MASK_FROM_INDEX(i) (1ULL << ((i) - 1))
+#define KFD_SMI_EVENT_MSG_SIZE	96
+
+enum KFD_MIGRATE_TRIGGERS {
+  KFD_MIGRATE_TRIGGER_PREFETCH = 1,
+  KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU,
+  KFD_MIGRATE_TRIGGER_PAGEFAULT_CPU,
+  KFD_MIGRATE_TRIGGER_TTM_EVICTION
+};
+
+enum KFD_QUEUE_EVICTION_TRIGGERS {
+  KFD_QUEUE_EVICTION_TRIGGER_SVM = 1,
+  KFD_QUEUE_EVICTION_TRIGGER_USERPTR,
+  KFD_QUEUE_EVICTION_TRIGGER_TTM,
+  KFD_QUEUE_EVICTION_TRIGGER_SUSPEND,
+  KFD_QUEUE_EVICTION_CRIU_CHECKPOINT,
+  KFD_QUEUE_EVICTION_CRIU_RESTORE
+};
+
+enum KFD_SVM_UNMAP_TRIGGERS {
+  KFD_SVM_UNMAP_TRIGGER_MMU_NOTIFY = 1,
+  KFD_SVM_UNMAP_TRIGGER_MMU_NOTIFY_MIGRATE,
+  KFD_SVM_UNMAP_TRIGGER_UNMAP_FROM_CPU
+};
 
 struct kfd_ioctl_smi_events_args {
   __u32 gpuid;    /* to KFD */

include/rocm_smi/rocm_smi.h

@@ -304,13 +304,20 @@ typedef struct {
  * Event notification event types
  */
 typedef enum {
-  RSMI_EVT_NOTIF_VMFAULT = KFD_SMI_EVENT_VMFAULT,  //!< VM page fault
+  RSMI_EVT_NOTIF_VMFAULT = 1,  //!< 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_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_MIGRATE_START = 5,
+  RSMI_EVT_NOTIF_MIGRATE_END = 6,
+  RSMI_EVT_NOTIF_PAGE_FAULT_START = 7,
+  RSMI_EVT_NOTIF_PAGE_FAULT_END = 8,
+  RSMI_EVT_NOTIF_QUEUE_EVICTION = 9,
+  RSMI_EVT_NOTIF_QUEUE_RESTORE = 10,
+  RSMI_EVT_NOTIF_UNMAP_FROM_GPU = 11,
+  RSMI_EVT_NOTIF_ALL_PROCESS = 64,
+  RSMI_EVT_NOTIF_LAST = RSMI_EVT_NOTIF_UNMAP_FROM_GPU
 } rsmi_evt_notification_type_t;
 
 /**
@@ -319,7 +326,29 @@ typedef enum {
 #define RSMI_EVENT_MASK_FROM_INDEX(i) (1ULL << ((i) - 1))
 
 //! Maximum number of characters an event notification message will be
-#define MAX_EVENT_NOTIFICATION_MSG_SIZE 64
+#define MAX_EVENT_NOTIFICATION_MSG_SIZE 96
+
+typedef enum {
+  RSMI_EVT_NOTIF_MIGRATE_TRIGGER_PREFETCH = 1,
+  RSMI_EVT_NOTIF_MIGRATE_TRIGGER_PAGEFAULT_GPU,
+  RSMI_EVT_NOTIF_MIGRATE_TRIGGER_PAGEFAULT_CPU,
+  RSMI_EVT_NOTIF_MIGRATE_TRIGGER_TTM_EVICTION
+} rsmi_evt_notification_migrate_trigger_type_t;
+
+typedef enum {
+  RSMI_EVT_NOTIF_QUEUE_EVICTION_TRIGGER_SVM = 1,
+  RSMI_EVT_NOTIF_QUEUE_EVICTION_TRIGGER_USERPTR,
+  RSMI_EVT_NOTIF_QUEUE_EVICTION_TRIGGER_TTM,
+  RSMI_EVT_NOTIF_QUEUE_EVICTION_TRIGGER_SUSPEND,
+  RSMI_EVT_NOTIF_QUEUE_EVICTION_CRIU_CHECKPOINT,
+  RSMI_EVT_NOTIF_QUEUE_EVICTION_CRIU_RESTORE
+} rsmi_evt_notification_queue_eviction_trigger_type_t;
+
+typedef enum {
+  RSMI_EVT_NOTIF_SVM_UNMAP_TRIGGER_MMU_NOTIFY = 1,
+  RSMI_EVT_NOTIF_SVM_UNMAP_TRIGGER_MMU_NOTIFY_MIGRATE,
+  RSMI_EVT_NOTIF_SVM_UNMAP_TRIGGER_UNMAP_FROM_CPU
+} rsmi_evt_notification_svm_unmap_trigger_type_t;
 
 /**
  * Event notification data returned from event notification API

src/rocm_smi.cc

@@ -4013,7 +4013,10 @@ rsmi_event_notification_get(int timeout_ms,
            reinterpret_cast<rsmi_evt_notification_data_t *>(&data[*num_elem]);
 
       uint32_t event;
-      while (fscanf(anon_fp, "%x %63s\n", &event,
+      #define __LEN__(X) #X
+      #define LEN(X) __LEN__(X)
+
+      while (fscanf(anon_fp, "%x %" LEN(MAX_EVENT_NOTIFICATION_MSG_SIZE)"[^\n]\n", &event,
                           reinterpret_cast<char *>(&data_item->message)) == 2) {
         /* Output is in format as "event information\n"
          * Both event are expressed in hex.