[PATCH v3 4/4] drm/amdkfd: SMI report dropped event count

[Philip Yang]

Add new SMI event to report the dropped event count.

When the event kfifo is full, drop count is not zero, or no enough space
left to store the event message, increase drop count.

After reading event out from kfifo, if event was dropped, drop_count is
not zero, generate a dropped event record and reset drop count to zero.

Signed-off-by: Philip Yang <Philip.Yang at amd.com>
---
 drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 25 +++++++++++++++++----
 include/uapi/linux/kfd_ioctl.h              |  6 +++++
 2 files changed, 27 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index 9b8169761ec5..db321b8ea127 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -42,6 +42,7 @@ struct kfd_smi_client {
 	struct rcu_head rcu;
 	pid_t pid;
 	bool suser;
+	u32 drop_count;
 };
 
 #define KFD_MAX_KFIFO_SIZE	8192
@@ -103,12 +104,26 @@ static ssize_t kfd_smi_ev_read(struct file *filep, char __user *user,
 	}
 	to_copy = min(size, to_copy);
 	ret = kfifo_out(&client->fifo, buf, to_copy);
-	spin_unlock(&client->lock);
 	if (ret <= 0) {
+		spin_unlock(&client->lock);
 		ret = -EAGAIN;
 		goto ret_err;
 	}
 
+	if (client->drop_count) {
+		char msg[KFD_SMI_EVENT_MSG_SIZE];
+		int len;
+
+		len = snprintf(msg, sizeof(msg), "%x ", KFD_SMI_EVENT_DROPPED_EVENT);
+		len += snprintf(msg + len, sizeof(msg) - len,
+				KFD_EVENT_FMT_DROPPED_EVENT(ktime_get_boottime_ns(),
+				client->pid, client->drop_count));
+		kfifo_in(&client->fifo, msg, len);
+		client->drop_count = 0;
+	}
+
+	spin_unlock(&client->lock);
+
 	ret = copy_to_user(user, buf, to_copy);
 	if (ret) {
 		ret = -EFAULT;
@@ -182,13 +197,15 @@ static void add_event_to_kfifo(pid_t pid, struct kfd_node *dev,
 	list_for_each_entry_rcu(client, &dev->smi_clients, list) {
 		if (!kfd_smi_ev_enabled(pid, client, smi_event))
 			continue;
+
 		spin_lock(&client->lock);
-		if (kfifo_avail(&client->fifo) >= len) {
+		if (!client->drop_count && kfifo_avail(&client->fifo) >= len) {
 			kfifo_in(&client->fifo, event_msg, len);
 			wake_up_all(&client->wait_queue);
 		} else {
-			pr_debug("smi_event(EventID: %u): no space left\n",
-					smi_event);
+			client->drop_count++;
+			pr_debug("smi_event(EventID: %u): no space left drop_count %d\n",
+				 smi_event, client->drop_count);
 		}
 		spin_unlock(&client->lock);
 	}
diff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.h
index fa9f9846b88e..7afd66d45313 100644
--- a/include/uapi/linux/kfd_ioctl.h
+++ b/include/uapi/linux/kfd_ioctl.h
@@ -530,6 +530,7 @@ enum kfd_smi_event {
 	KFD_SMI_EVENT_QUEUE_EVICTION = 9,
 	KFD_SMI_EVENT_QUEUE_RESTORE = 10,
 	KFD_SMI_EVENT_UNMAP_FROM_GPU = 11,
+	KFD_SMI_EVENT_DROPPED_EVENT = 12,
 
 	/*
 	 * max event number, as a flag bit to get events from all processes,
@@ -610,6 +611,7 @@ struct kfd_ioctl_smi_events_args {
  *    rw: 'W' for write page fault, 'R' for read page fault
  *    rescheduled: 'R' if the queue restore failed and rescheduled to try again
  *    error_code: migrate failure error code, 0 if no error
+ *    drop_count: how many events dropped when fifo is full
  */
 #define KFD_EVENT_FMT_UPDATE_GPU_RESET(reset_seq_num, reset_cause)\
 		"%x %s\n", (reset_seq_num), (reset_cause)
@@ -645,6 +647,10 @@ struct kfd_ioctl_smi_events_args {
 		"%lld -%d @%lx(%lx) %x %d\n", (ns), (pid), (addr), (size),\
 		(node), (unmap_trigger)
 
+#define KFD_EVENT_FMT_DROPPED_EVENT(ns, pid, drop_count)\
+		"%lld -%d %d\n", (ns), (pid), (drop_count)
+
+
 /**************************************************************************************************
  * CRIU IOCTLs (Checkpoint Restore In Userspace)
  *
-- 
2.43.2