bnxt_en: add an nvm test for hw diagnose

Add an NVM test function for devlink hw reporter.
In this function an NVM VPD area is read followed by
a write. Test result is cached and if it is successful then
the next test can be conducted only after HW_RETEST_MIN_TIME to
avoid frequent writes to the NVM.
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Vikas Gupta <vikas.gupta@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/broadcom/bnxt/bnxt.h

@@ -1548,13 +1548,25 @@ enum bnxt_hw_err {
 	BNXT_HW_STATUS_NVM_WRITE_ERR		= 0x1,
 	BNXT_HW_STATUS_NVM_ERASE_ERR		= 0x2,
 	BNXT_HW_STATUS_NVM_UNKNOWN_ERR		= 0x3,
+	BNXT_HW_STATUS_NVM_TEST_VPD_ENT_ERR	= 0x4,
+	BNXT_HW_STATUS_NVM_TEST_VPD_READ_ERR	= 0x5,
+	BNXT_HW_STATUS_NVM_TEST_VPD_WRITE_ERR	= 0x6,
+	BNXT_HW_STATUS_NVM_TEST_INCMPL_ERR	= 0x7,
 };
 
 struct bnxt_hw_health {
 	u32 nvm_err_address;
 	u32 nvm_write_errors;
 	u32 nvm_erase_errors;
+	u32 nvm_test_vpd_ent_errors;
+	u32 nvm_test_vpd_read_errors;
+	u32 nvm_test_vpd_write_errors;
+	u32 nvm_test_incmpl_errors;
 	u8 synd;
+	/* max a test in a day if previous test was successful */
+#define HW_RETEST_MIN_TIME	(1000 * 3600 * 24)
+	u8 nvm_test_result;
+	unsigned long nvm_test_timestamp;
 	struct devlink_health_reporter *hw_reporter;
 };
 

drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c

@@ -20,6 +20,7 @@
 #include "bnxt_ulp.h"
 #include "bnxt_ptp.h"
 #include "bnxt_coredump.h"
+#include "bnxt_nvm_defs.h"	/* NVRAM content constant and structure defs */
 
 static void __bnxt_fw_recover(struct bnxt *bp)
 {
@@ -263,20 +264,82 @@ static const char *hw_err_str(u8 synd)
 		return "nvm erase error";
 	case BNXT_HW_STATUS_NVM_UNKNOWN_ERR:
 		return "unrecognized nvm error";
+	case BNXT_HW_STATUS_NVM_TEST_VPD_ENT_ERR:
+		return "nvm test vpd entry error";
+	case BNXT_HW_STATUS_NVM_TEST_VPD_READ_ERR:
+		return "nvm test vpd read error";
+	case BNXT_HW_STATUS_NVM_TEST_VPD_WRITE_ERR:
+		return "nvm test vpd write error";
+	case BNXT_HW_STATUS_NVM_TEST_INCMPL_ERR:
+		return "nvm test incomplete error";
 	default:
 		return "unknown hw error";
 	}
 }
 
+static void bnxt_nvm_test(struct bnxt *bp)
+{
+	struct bnxt_hw_health *h = &bp->hw_health;
+	u32 datalen;
+	u16 index;
+	u8 *buf;
+
+	if (!h->nvm_test_result) {
+		if (!h->nvm_test_timestamp ||
+		    time_after(jiffies, h->nvm_test_timestamp +
+					msecs_to_jiffies(HW_RETEST_MIN_TIME)))
+			h->nvm_test_timestamp = jiffies;
+		else
+			return;
+	}
+
+	if (bnxt_find_nvram_item(bp->dev, BNX_DIR_TYPE_VPD,
+				 BNX_DIR_ORDINAL_FIRST, BNX_DIR_EXT_NONE,
+				 &index, NULL, &datalen) || !datalen) {
+		h->nvm_test_result = BNXT_HW_STATUS_NVM_TEST_VPD_ENT_ERR;
+		h->nvm_test_vpd_ent_errors++;
+		return;
+	}
+
+	buf = kzalloc(datalen, GFP_KERNEL);
+	if (!buf) {
+		h->nvm_test_result = BNXT_HW_STATUS_NVM_TEST_INCMPL_ERR;
+		h->nvm_test_incmpl_errors++;
+		return;
+	}
+
+	if (bnxt_get_nvram_item(bp->dev, index, 0, datalen, buf)) {
+		h->nvm_test_result = BNXT_HW_STATUS_NVM_TEST_VPD_READ_ERR;
+		h->nvm_test_vpd_read_errors++;
+		goto err;
+	}
+
+	if (bnxt_flash_nvram(bp->dev, BNX_DIR_TYPE_VPD, BNX_DIR_ORDINAL_FIRST,
+			     BNX_DIR_EXT_NONE, 0, 0, buf, datalen)) {
+		h->nvm_test_result = BNXT_HW_STATUS_NVM_TEST_VPD_WRITE_ERR;
+		h->nvm_test_vpd_write_errors++;
+	}
+
+err:
+	kfree(buf);
+}
+
 static int bnxt_hw_diagnose(struct devlink_health_reporter *reporter,
 			    struct devlink_fmsg *fmsg,
 			    struct netlink_ext_ack *extack)
 {
 	struct bnxt *bp = devlink_health_reporter_priv(reporter);
 	struct bnxt_hw_health *h = &bp->hw_health;
+	u8 synd = h->synd;
 	int rc;
 
-	rc = devlink_fmsg_string_pair_put(fmsg, "Status", hw_err_str(h->synd));
+	bnxt_nvm_test(bp);
+	if (h->nvm_test_result) {
+		synd = h->nvm_test_result;
+		devlink_health_report(h->hw_reporter, hw_err_str(synd), NULL);
+	}
+
+	rc = devlink_fmsg_string_pair_put(fmsg, "Status", hw_err_str(synd));
 	if (rc)
 		return rc;
 	rc = devlink_fmsg_u32_pair_put(fmsg, "nvm_write_errors", h->nvm_write_errors);
@@ -285,6 +348,23 @@ static int bnxt_hw_diagnose(struct devlink_health_reporter *reporter,
 	rc = devlink_fmsg_u32_pair_put(fmsg, "nvm_erase_errors", h->nvm_erase_errors);
 	if (rc)
 		return rc;
+	rc = devlink_fmsg_u32_pair_put(fmsg, "nvm_test_vpd_ent_errors",
+				       h->nvm_test_vpd_ent_errors);
+	if (rc)
+		return rc;
+	rc = devlink_fmsg_u32_pair_put(fmsg, "nvm_test_vpd_read_errors",
+				       h->nvm_test_vpd_read_errors);
+	if (rc)
+		return rc;
+	rc = devlink_fmsg_u32_pair_put(fmsg, "nvm_test_vpd_write_errors",
+				       h->nvm_test_vpd_write_errors);
+	if (rc)
+		return rc;
+	rc = devlink_fmsg_u32_pair_put(fmsg, "nvm_test_incomplete_errors",
+				       h->nvm_test_incmpl_errors);
+	if (rc)
+		return rc;
+
 	return 0;
 }
 

drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c

@@ -2168,11 +2168,7 @@ static void bnxt_print_admin_err(struct bnxt *bp)
 	netdev_info(bp->dev, "PF does not have admin privileges to flash or reset the device\n");
 }
 
-static int bnxt_find_nvram_item(struct net_device *dev, u16 type, u16 ordinal,
-				u16 ext, u16 *index, u32 *item_length,
-				u32 *data_length);
-
-static int bnxt_flash_nvram(struct net_device *dev, u16 dir_type,
+int bnxt_flash_nvram(struct net_device *dev, u16 dir_type,
 		     u16 dir_ordinal, u16 dir_ext, u16 dir_attr,
 		     u32 dir_item_len, const u8 *data,
 		     size_t data_len)
@@ -2819,7 +2815,7 @@ static int bnxt_get_nvram_directory(struct net_device *dev, u32 len, u8 *data)
 	return rc;
 }
 
-static int bnxt_get_nvram_item(struct net_device *dev, u32 index, u32 offset,
+int bnxt_get_nvram_item(struct net_device *dev, u32 index, u32 offset,
 			u32 length, u8 *data)
 {
 	struct bnxt *bp = netdev_priv(dev);
@@ -2854,7 +2850,7 @@ static int bnxt_get_nvram_item(struct net_device *dev, u32 index, u32 offset,
 	return rc;
 }
 
-static int bnxt_find_nvram_item(struct net_device *dev, u16 type, u16 ordinal,
+int bnxt_find_nvram_item(struct net_device *dev, u16 type, u16 ordinal,
 			 u16 ext, u16 *index, u32 *item_length,
 			 u32 *data_length)
 {

drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.h

@@ -56,6 +56,13 @@ int bnxt_hwrm_firmware_reset(struct net_device *dev, u8 proc_type,
 int bnxt_flash_package_from_fw_obj(struct net_device *dev, const struct firmware *fw,
 				   u32 install_type);
 int bnxt_get_pkginfo(struct net_device *dev, char *ver, int size);
+int bnxt_find_nvram_item(struct net_device *dev, u16 type, u16 ordinal, u16 ext,
+			 u16 *index, u32 *item_length, u32 *data_length);
+int bnxt_get_nvram_item(struct net_device *dev, u32 index, u32 offset,
+			u32 length, u8 *data);
+int bnxt_flash_nvram(struct net_device *dev, u16 dir_type, u16 dir_ordinal,
+		     u16 dir_ext, u16 dir_attr, u32 dir_item_len,
+		     const u8 *data, size_t data_len);
 void bnxt_ethtool_init(struct bnxt *bp);
 void bnxt_ethtool_free(struct bnxt *bp);