habanalabs: expose reset counters via existing INFO IOCTL

Expose both soft and hard reset counts via INFO IOCTL.
This will allow system management applications to easily check
if the device has undergone reset.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

drivers/misc/habanalabs/habanalabs_ioctl.c

@@ -242,6 +242,22 @@ static int get_clk_rate(struct hl_device *hdev, struct hl_info_args *args)
 		min((size_t) max_size, sizeof(clk_rate))) ? -EFAULT : 0;
 }
 
+static int get_reset_count(struct hl_device *hdev, struct hl_info_args *args)
+{
+	struct hl_info_reset_count reset_count = {0};
+	u32 max_size = args->return_size;
+	void __user *out = (void __user *) (uintptr_t) args->return_pointer;
+
+	if ((!max_size) || (!out))
+		return -EINVAL;
+
+	reset_count.hard_reset_cnt = hdev->hard_reset_cnt;
+	reset_count.soft_reset_cnt = hdev->soft_reset_cnt;
+
+	return copy_to_user(out, &reset_count,
+		min((size_t) max_size, sizeof(reset_count))) ? -EFAULT : 0;
+}
+
 static int _hl_info_ioctl(struct hl_fpriv *hpriv, void *data,
 				struct device *dev)
 {
@@ -260,6 +276,9 @@ static int _hl_info_ioctl(struct hl_fpriv *hpriv, void *data,
 	case HL_INFO_DEVICE_STATUS:
 		return device_status_info(hdev, args);
 
+	case HL_INFO_RESET_COUNT:
+		return get_reset_count(hdev, args);
+
 	default:
 		break;
 	}

include/uapi/misc/habanalabs.h

@@ -98,6 +98,9 @@ enum hl_device_status {
  * HL_INFO_CLK_RATE            - Retrieve the current and maximum clock rate
  *                               of the device in MHz. The maximum clock rate is
  *                               configurable via sysfs parameter
+ * HL_INFO_RESET_COUNT   - Retrieve the counts of the soft and hard reset
+ *                         operations performed on the device since the last
+ *                         time the driver was loaded.
  */
 #define HL_INFO_HW_IP_INFO		0
 #define HL_INFO_HW_EVENTS		1
@@ -107,6 +110,7 @@ enum hl_device_status {
 #define HL_INFO_DEVICE_UTILIZATION	6
 #define HL_INFO_HW_EVENTS_AGGREGATE	7
 #define HL_INFO_CLK_RATE		8
+#define HL_INFO_RESET_COUNT		9
 
 #define HL_INFO_VERSION_MAX_LEN	128
 #define HL_INFO_CARD_NAME_MAX_LEN	16
@@ -160,6 +164,11 @@ struct hl_info_clk_rate {
 	__u32 max_clk_rate_mhz;
 };
 
+struct hl_info_reset_count {
+	__u32 hard_reset_cnt;
+	__u32 soft_reset_cnt;
+};
+
 struct hl_info_args {
 	/* Location of relevant struct in userspace */
 	__u64 return_pointer;