drive-health-metrics/smart_json.go

package main

import (
	"fmt"
	"strings"
)

// wearAttr maps vendor SSD-life attribute IDs to a source label. For all of
// these the normalized VALUE expresses "% life remaining".
var wearAttr = []struct {
	id  int
	src string
}{
	{173, "Micron/ID173"},
	{202, "Intel/ID202"},
	{231, "Intel/ID231"},
	{177, "Samsung/ID177"},
	{233, "Generic/ID233"},
}

// ataAttr is one parsed row of an ATA SMART attribute table, shared by the JSON
// and text paths. raw is the attribute's raw counter; value/worst are the
// vendor-normalized current/worst values.
type ataAttr struct {
	value      *int
	worst      *int
	whenFailed string
	raw        *int
}

// attrRaw returns the raw counter for attribute id, or nil when absent.
func attrRaw(attrs map[int]ataAttr, id int) *int {
	if a, ok := attrs[id]; ok {
		return a.raw
	}
	return nil
}

// attrVal returns the normalized current value for attribute id, or nil.
func attrVal(attrs map[int]ataAttr, id int) *int {
	if a, ok := attrs[id]; ok {
		return a.value
	}
	return nil
}

// attrsFailed reports whether any attribute is flagged failed now or in the
// past — the basis for the PASSED_BY_ATTR/FAILED verdict when no explicit
// overall health result is available.
func attrsFailed(attrs map[int]ataAttr) bool {
	for _, a := range attrs {
		wf := strings.ToLower(a.whenFailed)
		if wf == "now" || wf == "past" {
			return true
		}
	}
	return false
}

// applyAtaCounters maps the parsed ATA attribute table onto the defect, wear,
// reserve-block, and host-write fields shared by the JSON and text paths. Power-
// on hours fall back to attribute 9 only when not already set from a dedicated
// field. Path-specific fallbacks (power-cycle/temperature on text, SCSI/NVMe on
// JSON) stay with their callers.
func applyAtaCounters(attrs map[int]ataAttr, d *Drive) {
	d.Reallocated = attrRaw(attrs, 5)
	d.ReallocatedEvents = attrRaw(attrs, 196)
	d.Pending = attrRaw(attrs, 197)
	d.Uncorrectable = attrRaw(attrs, 198)
	d.UdmaCrc = attrRaw(attrs, 199)
	d.ReportedUncorrect = attrRaw(attrs, 187)
	d.RuntimeBadBlocks = attrRaw(attrs, 183)
	d.EndToEnd = attrRaw(attrs, 184)
	if d.PowerOnHours == nil {
		d.PowerOnHours = attrRaw(attrs, 9)
	}

	// Wear (vendor-normalized; VALUE = % remaining).
	for _, w := range wearAttr {
		if v := attrVal(attrs, w.id); v != nil {
			d.WearPctRemaining = v
			if a, ok := attrs[w.id]; ok {
				d.WearPctWorst = a.worst
			}
			d.WearSrc = w.src
			d.WearPctConsumed = pInt(100 - *v)
			break
		}
	}

	// Micron ID180 reserve blocks (VALUE = % remaining) and ID246 host writes.
	if a, ok := attrs[180]; ok {
		d.UnusedReservePct = a.value
	}
	if lba := attrRaw(attrs, 246); lba != nil && *lba > 0 {
		d.HostWrittenTB = pF(float64(*lba) * 512.0 / 1e12)
	}
}

// parseSmartJSON fills d from a smartctl -j object (ATA/SATA, SAS/SCSI, or NVMe).
func parseSmartJSON(j map[string]interface{}, d *Drive) {
	if j == nil {
		return
	}

	d.Model = first(jStr(j, "model_name"), jStr(j, "scsi_model_name"))
	d.Serial = jStr(j, "serial_number")
	d.Transport = jStr(j, "scsi_transport_protocol", "name")
	d.Firmware = first(jStr(j, "firmware_version"), jStr(j, "scsi_revision"), jStr(j, "revision"))

	if cap := jInt(j, "user_capacity", "bytes"); cap != nil && *cap > 0 {
		d.Capacity = fmt.Sprintf("%.2f TB", float64(*cap)/1e12)
	}

	switch rr := jInt(j, "rotation_rate"); {
	case rr != nil && *rr == 0:
		d.Rotation = "SSD"
	case rr != nil:
		d.Rotation = fmt.Sprintf("%d rpm", *rr)
	default:
		d.Rotation = "SSD" // Absent rotation_rate: assume SSD; NVMe is corrected just below.
	}
	if strings.Contains(strings.ToLower(jStr(j, "device", "type")), "nvme") ||
		jObj(j, "nvme_smart_health_information_log") != nil {
		d.Rotation = "NVMe"
	}

	d.PowerOnHours = jInt(j, "power_on_time", "hours")
	d.PowerCycleCount = jInt(j, "power_cycle_count")
	d.TempC = jInt(j, "temperature", "current")

	// ---- ATA attribute table ----
	attrs := map[int]ataAttr{}
	if table, ok := jLeaf(j, "ata_smart_attributes", "table").([]interface{}); ok {
		for _, it := range table {
			a, ok := it.(map[string]interface{})
			if !ok {
				continue
			}
			id := jInt(a, "id")
			if id == nil {
				continue
			}
			at := ataAttr{
				value:      jInt(a, "value"),
				worst:      jInt(a, "worst"),
				whenFailed: jStr(a, "when_failed"),
			}
			// Prefer the leading integer of raw.string (raw.value overflows
			// for some attributes); fall back to raw.value.
			if rs := jStr(a, "raw", "string"); rs != "" {
				if n, ok := firstInt(rs); ok {
					at.raw = &n
				}
			}
			if at.raw == nil {
				at.raw = jInt(a, "raw", "value")
			}
			attrs[*id] = at
		}
	}

	// ---- SMART health verdict ----
	if passed := jBoolPtr(j, "smart_status", "passed"); passed != nil {
		if *passed {
			d.SmartHealth = "PASSED"
		} else {
			d.SmartHealth = "FAILED"
		}
	} else if len(attrs) > 0 {
		if attrsFailed(attrs) {
			d.SmartHealth = "FAILED"
		} else {
			d.SmartHealth = "PASSED_BY_ATTR"
		}
	} else {
		d.SmartHealth = "UNKNOWN"
	}

	// Defect, wear, reserve, and host-write fields shared with the text path.
	applyAtaCounters(attrs, d)

	// ---- SCSI/SAS endurance + grown defect list ----
	if d.WearPctRemaining == nil {
		if pu := jInt(j, "scsi_percentage_used_endurance_indicator"); pu != nil {
			d.WearPctConsumed = pu
			d.WearPctRemaining = pInt(100 - *pu)
			d.WearSrc = "SCSI/endurance"
		}
	}
	if grown := jInt(j, "scsi_grown_defect_list"); grown != nil && d.Reallocated == nil {
		d.Reallocated = grown
	}
	// SAS drives have no ATA attribute table; their hard-defect signals live in
	// the SCSI logs. Map them onto the fields the scorer already grades: total
	// uncorrected read/write/verify errors -> uncorrectable sectors, and the
	// pending (to-be-reassigned) defect count -> current pending sectors.
	if d.Uncorrectable == nil {
		if ec := jObj(j, "scsi_error_counter_log"); ec != nil {
			sum, any := 0, false
			for _, op := range []string{"read", "write", "verify"} {
				if u := jInt(ec, op, "total_uncorrected_errors"); u != nil {
					any = true
					sum += *u
				}
			}
			if any {
				d.Uncorrectable = pInt(sum)
			}
		}
	}
	if d.Pending == nil {
		if pd := jInt(j, "scsi_pending_defects", "count"); pd != nil {
			d.Pending = pd
		}
	}

	// ---- NVMe SMART/Health log ----
	if nv := jObj(j, "nvme_smart_health_information_log"); nv != nil {
		d.NvmeCriticalWarning = jInt(nv, "critical_warning")
		d.NvmeAvailSpare = jInt(nv, "available_spare")
		d.NvmeAvailSpareThresh = jInt(nv, "available_spare_threshold")
		d.NvmeMediaErrors = jInt(nv, "media_errors")
		if d.PowerOnHours == nil {
			d.PowerOnHours = jInt(nv, "power_on_hours")
		}
		if d.PowerCycleCount == nil {
			d.PowerCycleCount = jInt(nv, "power_cycles")
		}
		if d.TempC == nil {
			d.TempC = jInt(nv, "temperature")
		}
		if pu := jInt(nv, "percentage_used"); pu != nil {
			d.WearPctConsumed = pu
			d.WearPctRemaining = pInt(100 - *pu)
			d.WearSrc = "NVMe/percentage_used"
		}
		// Treat NVMe media+data integrity errors like uncorrectable sectors.
		if d.Uncorrectable == nil && d.NvmeMediaErrors != nil {
			d.Uncorrectable = d.NvmeMediaErrors
		}
	}

	// Capture smartctl passthrough diagnostics.
	if msgs, ok := jLeaf(j, "smartctl", "messages").([]interface{}); ok {
		var parts []string
		for _, mm := range msgs {
			if mo, ok := mm.(map[string]interface{}); ok {
				if s := jStr(mo, "string"); s != "" {
					parts = append(parts, s)
				}
			}
		}
		d.SmartctlMessages = strings.Join(parts, "; ")
	}
}

// jsonUsable reports whether the parsed object carries real identity + health.
func jsonUsable(j map[string]interface{}) bool {
	if j == nil {
		return false
	}
	hasID := jStr(j, "model_name") != "" || jStr(j, "scsi_model_name") != ""
	if !hasID {
		return false
	}
	if jObj(j, "ata_smart_attributes") != nil ||
		jObj(j, "smart_status") != nil ||
		jObj(j, "nvme_smart_health_information_log") != nil ||
		jInt(j, "scsi_percentage_used_endurance_indicator") != nil {
		return true
	}
	return false
}

// first returns the first non-empty string in vals, or "".
func first(vals ...string) string {
	for _, v := range vals {
		if v != "" {
			return v
		}
	}
	return ""
}