hinic: add support to handle hw abnormal event

add support to handle hw abnormal event such as hardware failure,
cable unplugged,link error

Signed-off-by: Luo bin <luobin9@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 283 insertions, 3 deletions

diff --git a/drivers/net/ethernet/huawei/hinic/hinic_devlink.c b/drivers/net/ethernet/huawei/hinic/hinic_devlink.c
index a40a10ac1ee9..c6adc776f3c8 100644
--- a/drivers/net/ethernet/huawei/hinic/hinic_devlink.c
+++ b/drivers/net/ethernet/huawei/hinic/hinic_devlink.c
@@ -16,9 +16,9 @@
 #include <net/devlink.h>
 #include <linux/firmware.h>
 
-#include "hinic_dev.h"
 #include "hinic_port.h"
 #include "hinic_devlink.h"
+#include "hinic_hw_dev.h"
 
 static bool check_image_valid(struct hinic_devlink_priv *priv, const u8 *buf,
 			      u32 image_size, struct host_image_st *host_image)
@@ -317,12 +317,292 @@ void hinic_devlink_free(struct devlink *devlink)
 	devlink_free(devlink);
 }
 
-int hinic_devlink_register(struct devlink *devlink, struct device *dev)
+int hinic_devlink_register(struct hinic_devlink_priv *priv, struct device *dev)
 {
+	struct devlink *devlink = priv_to_devlink(priv);
+
 	return devlink_register(devlink, dev);
 }
 
-void hinic_devlink_unregister(struct devlink *devlink)
+void hinic_devlink_unregister(struct hinic_devlink_priv *priv)
 {
+	struct devlink *devlink = priv_to_devlink(priv);
+
 	devlink_unregister(devlink);
 }
+
+static int chip_fault_show(struct devlink_fmsg *fmsg,
+			   struct hinic_fault_event *event)
+{
+	char fault_level[FAULT_TYPE_MAX][FAULT_SHOW_STR_LEN + 1] = {
+		"fatal", "reset", "flr", "general", "suggestion"};
+	char level_str[FAULT_SHOW_STR_LEN + 1] = {0};
+	u8 level;
+	int err;
+
+	level = event->event.chip.err_level;
+	if (level < FAULT_LEVEL_MAX)
+		strncpy(level_str, fault_level[level], strlen(fault_level[level]));
+	else
+		strncpy(level_str, "Unknown", strlen("Unknown"));
+
+	if (level == FAULT_LEVEL_SERIOUS_FLR) {
+		err = devlink_fmsg_u32_pair_put(fmsg, "Function level err func_id",
+						(u32)event->event.chip.func_id);
+		if (err)
+			return err;
+	}
+
+	err = devlink_fmsg_u8_pair_put(fmsg, "module_id", event->event.chip.node_id);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "err_type", (u32)event->event.chip.err_type);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_string_pair_put(fmsg, "err_level", level_str);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "err_csr_addr",
+					event->event.chip.err_csr_addr);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "err_csr_value",
+					event->event.chip.err_csr_value);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int fault_report_show(struct devlink_fmsg *fmsg,
+			     struct hinic_fault_event *event)
+{
+	char fault_type[FAULT_TYPE_MAX][FAULT_SHOW_STR_LEN + 1] = {
+		"chip", "ucode", "mem rd timeout", "mem wr timeout",
+		"reg rd timeout", "reg wr timeout", "phy fault"};
+	char type_str[FAULT_SHOW_STR_LEN + 1] = {0};
+	int err;
+
+	if (event->type < FAULT_TYPE_MAX)
+		strncpy(type_str, fault_type[event->type], strlen(fault_type[event->type]));
+	else
+		strncpy(type_str, "Unknown", strlen("Unknown"));
+
+	err = devlink_fmsg_string_pair_put(fmsg, "Fault type", type_str);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_binary_pair_put(fmsg, "Fault raw data",
+					   event->event.val, sizeof(event->event.val));
+	if (err)
+		return err;
+
+	switch (event->type) {
+	case FAULT_TYPE_CHIP:
+		err = chip_fault_show(fmsg, event);
+		if (err)
+			return err;
+		break;
+	case FAULT_TYPE_UCODE:
+		err = devlink_fmsg_u8_pair_put(fmsg, "Cause_id", event->event.ucode.cause_id);
+		if (err)
+			return err;
+		err = devlink_fmsg_u8_pair_put(fmsg, "core_id", event->event.ucode.core_id);
+		if (err)
+			return err;
+		err = devlink_fmsg_u8_pair_put(fmsg, "c_id", event->event.ucode.c_id);
+		if (err)
+			return err;
+		err = devlink_fmsg_u8_pair_put(fmsg, "epc", event->event.ucode.epc);
+		if (err)
+			return err;
+		break;
+	case FAULT_TYPE_MEM_RD_TIMEOUT:
+	case FAULT_TYPE_MEM_WR_TIMEOUT:
+		err = devlink_fmsg_u32_pair_put(fmsg, "Err_csr_ctrl",
+						event->event.mem_timeout.err_csr_ctrl);
+		if (err)
+			return err;
+		err = devlink_fmsg_u32_pair_put(fmsg, "err_csr_data",
+						event->event.mem_timeout.err_csr_data);
+		if (err)
+			return err;
+		err = devlink_fmsg_u32_pair_put(fmsg, "ctrl_tab",
+						event->event.mem_timeout.ctrl_tab);
+		if (err)
+			return err;
+		err = devlink_fmsg_u32_pair_put(fmsg, "mem_index",
+						event->event.mem_timeout.mem_index);
+		if (err)
+			return err;
+		break;
+	case FAULT_TYPE_REG_RD_TIMEOUT:
+	case FAULT_TYPE_REG_WR_TIMEOUT:
+		err = devlink_fmsg_u32_pair_put(fmsg, "Err_csr", event->event.reg_timeout.err_csr);
+		if (err)
+			return err;
+		break;
+	case FAULT_TYPE_PHY_FAULT:
+		err = devlink_fmsg_u8_pair_put(fmsg, "Op_type", event->event.phy_fault.op_type);
+		if (err)
+			return err;
+		err = devlink_fmsg_u8_pair_put(fmsg, "port_id", event->event.phy_fault.port_id);
+		if (err)
+			return err;
+		err = devlink_fmsg_u8_pair_put(fmsg, "dev_ad", event->event.phy_fault.dev_ad);
+		if (err)
+			return err;
+
+		err = devlink_fmsg_u32_pair_put(fmsg, "csr_addr", event->event.phy_fault.csr_addr);
+		if (err)
+			return err;
+		err = devlink_fmsg_u32_pair_put(fmsg, "op_data", event->event.phy_fault.op_data);
+		if (err)
+			return err;
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int hinic_hw_reporter_dump(struct devlink_health_reporter *reporter,
+				  struct devlink_fmsg *fmsg, void *priv_ctx,
+				  struct netlink_ext_ack *extack)
+{
+	if (priv_ctx)
+		return fault_report_show(fmsg, priv_ctx);
+
+	return 0;
+}
+
+static int mgmt_watchdog_report_show(struct devlink_fmsg *fmsg,
+				     struct hinic_mgmt_watchdog_info *watchdog_info)
+{
+	int err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "Mgmt deadloop time_h", watchdog_info->curr_time_h);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "time_l", watchdog_info->curr_time_l);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "task_id", watchdog_info->task_id);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "sp", watchdog_info->sp);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "stack_current_used", watchdog_info->curr_used);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "peak_used", watchdog_info->peak_used);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "\n Overflow_flag", watchdog_info->is_overflow);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "stack_top", watchdog_info->stack_top);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "stack_bottom", watchdog_info->stack_bottom);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "mgmt_pc", watchdog_info->pc);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "lr", watchdog_info->lr);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_u32_pair_put(fmsg, "cpsr", watchdog_info->cpsr);
+	if (err)
+		return err;
+
+	err = devlink_fmsg_binary_pair_put(fmsg, "Mgmt register info",
+					   watchdog_info->reg, sizeof(watchdog_info->reg));
+	if (err)
+		return err;
+
+	err = devlink_fmsg_binary_pair_put(fmsg, "Mgmt dump stack(start from sp)",
+					   watchdog_info->data, sizeof(watchdog_info->data));
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int hinic_fw_reporter_dump(struct devlink_health_reporter *reporter,
+				  struct devlink_fmsg *fmsg, void *priv_ctx,
+				  struct netlink_ext_ack *extack)
+{
+	if (priv_ctx)
+		return mgmt_watchdog_report_show(fmsg, priv_ctx);
+
+	return 0;
+}
+
+static const struct devlink_health_reporter_ops hinic_hw_fault_reporter_ops = {
+	.name = "hw",
+	.dump = hinic_hw_reporter_dump,
+};
+
+static const struct devlink_health_reporter_ops hinic_fw_fault_reporter_ops = {
+	.name = "fw",
+	.dump = hinic_fw_reporter_dump,
+};
+
+int hinic_health_reporters_create(struct hinic_devlink_priv *priv)
+{
+	struct devlink *devlink = priv_to_devlink(priv);
+
+	priv->hw_fault_reporter =
+		devlink_health_reporter_create(devlink, &hinic_hw_fault_reporter_ops,
+					       0, priv);
+	if (IS_ERR(priv->hw_fault_reporter)) {
+		dev_warn(&priv->hwdev->hwif->pdev->dev, "Failed to create hw fault reporter, err: %ld\n",
+			 PTR_ERR(priv->hw_fault_reporter));
+		return PTR_ERR(priv->hw_fault_reporter);
+	}
+
+	priv->fw_fault_reporter =
+		devlink_health_reporter_create(devlink, &hinic_fw_fault_reporter_ops,
+					       0, priv);
+	if (IS_ERR(priv->fw_fault_reporter)) {
+		dev_warn(&priv->hwdev->hwif->pdev->dev, "Failed to create fw fault reporter, err: %ld\n",
+			 PTR_ERR(priv->fw_fault_reporter));
+		devlink_health_reporter_destroy(priv->hw_fault_reporter);
+		priv->hw_fault_reporter = NULL;
+		return PTR_ERR(priv->fw_fault_reporter);
+	}
+
+	return 0;
+}
+
+void hinic_health_reporters_destroy(struct hinic_devlink_priv *priv)
+{
+	if (!IS_ERR_OR_NULL(priv->fw_fault_reporter)) {
+		devlink_health_reporter_destroy(priv->fw_fault_reporter);
+		priv->fw_fault_reporter = NULL;
+	}
+
+	if (!IS_ERR_OR_NULL(priv->hw_fault_reporter)) {
+		devlink_health_reporter_destroy(priv->hw_fault_reporter);
+		priv->hw_fault_reporter = NULL;
+	}
+}