Add GPU serial number extraction for NVIDIA diagnostics

Parse inventory/output.log to extract GPU serial numbers from lspci output,
expose them via serials API, and add GPU category to web UI.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

internal/parser/vendors/nvidia/inventory_log.go

@@ -0,0 +1,143 @@
+package nvidia
+
+import (
+	"bufio"
+	"regexp"
+	"strings"
+
+	"git.mchus.pro/mchus/logpile/internal/models"
+	"git.mchus.pro/mchus/logpile/internal/parser"
+)
+
+var (
+	// Regex to extract GPU serial numbers from lspci output
+	// Example: "	Capabilities: [2f0 v1] Device Serial Number 14-17-dc-65-77-2d-b0-48"
+	gpuSerialRegex = regexp.MustCompile(`Device Serial Number\s+([\da-fA-F-]+)`)
+
+	// Regex to extract PCI BDF from lspci header
+	// Example: "2a:00.0 3D controller: NVIDIA Corporation Device 2335 (rev a1)"
+	// Note: lspci format is bus:device.function (e.g., "2a:00.0")
+	pciBDFRegex = regexp.MustCompile(`^([0-9a-fA-F]{2,4}:[0-9a-fA-F]{2}\.[0-9])\s+3D controller.*NVIDIA`)
+
+	// Regex to extract devname mappings from fieldiag command line
+	// Example: "devname=0000:ba:00.0,SXM5_SN_1653925027099"
+	devnameRegex = regexp.MustCompile(`devname=([\da-fA-F:\.]+),(\w+)`)
+)
+
+// ParseInventoryLog parses inventory/output.log to extract GPU serial numbers
+// from lspci output and map them to slots
+func ParseInventoryLog(content []byte, result *models.AnalysisResult) error {
+	if result.Hardware == nil || len(result.Hardware.GPUs) == 0 {
+		// No GPUs to update
+		return nil
+	}
+
+	scanner := bufio.NewScanner(strings.NewReader(string(content)))
+
+	// First pass: build mapping of PCI BDF -> Slot name from fieldiag command line
+	pciToSlot := make(map[string]string)
+	for scanner.Scan() {
+		line := scanner.Text()
+		// Look for fieldiag command with devname parameters
+		if strings.Contains(line, "devname=") && strings.Contains(line, "fieldiag") {
+			matches := devnameRegex.FindAllStringSubmatch(line, -1)
+			for _, match := range matches {
+				if len(match) == 3 {
+					pciBDF := match[1]
+					slotName := match[2]
+					// Extract slot number from name like "SXM5_SN_1653925027099"
+					// We want to map to slot like "GPUSXM5"
+					if strings.HasPrefix(slotName, "SXM") {
+						parts := strings.Split(slotName, "_")
+						if len(parts) >= 1 {
+							// Convert "SXM5" to "GPUSXM5"
+							slot := "GPU" + parts[0]
+							pciToSlot[pciBDF] = slot
+						}
+					}
+				}
+			}
+		}
+	}
+
+	// Second pass: extract GPU serial numbers from lspci output
+	scanner = bufio.NewScanner(strings.NewReader(string(content)))
+	var currentPCIBDF string
+	var currentSlot string
+
+	for scanner.Scan() {
+		line := scanner.Text()
+
+		// Check if this is a new GPU device header
+		if match := pciBDFRegex.FindStringSubmatch(line); len(match) > 1 {
+			currentPCIBDF = match[1]
+			// Normalize BDF format - lspci uses short format (bus:device.function)
+			// but fieldiag uses full format (domain:bus:device.function)
+			// Convert "2a:00.0" to "0000:2a:00.0"
+			normalizedBDF := currentPCIBDF
+			if len(strings.Split(currentPCIBDF, ":")) == 2 {
+				// Short format without domain, add 0000:
+				normalizedBDF = "0000:" + currentPCIBDF
+			}
+
+			// Map to slot name if we have it
+			if slot, ok := pciToSlot[normalizedBDF]; ok {
+				currentSlot = slot
+			} else if slot, ok := pciToSlot[currentPCIBDF]; ok {
+				currentSlot = slot
+			} else {
+				currentSlot = ""
+			}
+			continue
+		}
+
+		// Look for Device Serial Number in capabilities
+		if match := gpuSerialRegex.FindStringSubmatch(line); len(match) > 1 && currentSlot != "" {
+			serialNumber := match[1]
+			// Format: 14-17-dc-65-77-2d-b0-48
+			// Convert to more readable format: 48:b0:2d:77:65:dc:17:14 (reversed)
+			serialFormatted := formatGPUSerial(serialNumber)
+
+			// Find the GPU in our results and update its serial number
+			for i := range result.Hardware.GPUs {
+				if result.Hardware.GPUs[i].Slot == currentSlot {
+					result.Hardware.GPUs[i].SerialNumber = serialFormatted
+					break
+				}
+			}
+		}
+	}
+
+	return scanner.Err()
+}
+
+// formatGPUSerial formats GPU serial number from PCIe format to human-readable
+// Input: "14-17-dc-65-77-2d-b0-48" (little-endian from PCIe)
+// Output: "48:b0:2d:77:65:dc:17:14" (reversed to match GPU label)
+func formatGPUSerial(serial string) string {
+	parts := strings.Split(serial, "-")
+	if len(parts) != 8 {
+		return serial // Return as-is if unexpected format
+	}
+
+	// Reverse the bytes (PCIe reports in little-endian)
+	reversed := make([]string, len(parts))
+	for i := range parts {
+		reversed[len(parts)-1-i] = strings.ToUpper(parts[i])
+	}
+
+	return strings.Join(reversed, ":")
+}
+
+// findInventoryOutputLog finds the inventory/output.log file
+func findInventoryOutputLog(files []parser.ExtractedFile) *parser.ExtractedFile {
+	for _, f := range files {
+		// Look for inventory/output.log
+		path := strings.ToLower(f.Path)
+		if strings.Contains(path, "inventory/output.log") ||
+			strings.Contains(path, "inventory\\output.log") {
+			return &f
+		}
+	}
+	return nil
+}

internal/parser/vendors/nvidia/inventory_log_test.go

@@ -0,0 +1,87 @@
+package nvidia
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"git.mchus.pro/mchus/logpile/internal/parser"
+)
+
+func TestParseInventoryLog(t *testing.T) {
+	// Test with the real archive
+	archivePath := filepath.Join("../../../../example", "A514359X5A09844_logs-20260115-151707.tar")
+
+	// Check if file exists
+	if _, err := os.Stat(archivePath); os.IsNotExist(err) {
+		t.Skip("Test archive not found, skipping test")
+	}
+
+	// Extract files from archive
+	files, err := parser.ExtractArchive(archivePath)
+	if err != nil {
+		t.Fatalf("Failed to extract archive: %v", err)
+	}
+
+	// Find inventory/output.log
+	var inventoryLog *parser.ExtractedFile
+	for _, f := range files {
+		if strings.Contains(f.Path, "inventory/output.log") {
+			inventoryLog = &f
+			break
+		}
+	}
+
+	if inventoryLog == nil {
+		t.Fatal("inventory/output.log not found")
+	}
+
+	content := string(inventoryLog.Content)
+
+	// Test devname regex
+	t.Log("Testing devname extraction:")
+	lines := strings.Split(content, "\n")
+	for i, line := range lines {
+		if strings.Contains(line, "devname=") && strings.Contains(line, "fieldiag") {
+			t.Logf("Line %d: Found fieldiag command", i)
+			matches := devnameRegex.FindAllStringSubmatch(line, -1)
+			t.Logf("  Found %d devname matches", len(matches))
+			for _, match := range matches {
+				if len(match) == 3 {
+					t.Logf("    PCI: %s -> Slot: %s", match[1], match[2])
+				}
+			}
+			break
+		}
+	}
+
+	// Test lspci regex
+	t.Log("\nTesting lspci BDF extraction:")
+	serialCount := 0
+	bdfCount := 0
+	for i, line := range lines {
+		// Check for lines that look like lspci headers
+		if strings.Contains(line, "3D controller") && strings.Contains(line, "NVIDIA") {
+			t.Logf("Line %d: Potential lspci line: %q (starts with: %q)", i, line[:min(80, len(line))], line[:min(10, len(line))])
+			if match := pciBDFRegex.FindStringSubmatch(line); len(match) > 1 {
+				bdfCount++
+				t.Logf("  -> Matched BDF: %s", match[1])
+			} else {
+				t.Logf("  -> NO MATCH")
+			}
+		}
+		if match := gpuSerialRegex.FindStringSubmatch(line); len(match) > 1 {
+			serialCount++
+		}
+	}
+	t.Logf("\nTotal BDFs found: %d", bdfCount)
+	t.Logf("Total serials found: %d", serialCount)
+}
+
+func min(a, b int) int {
+	if a < b {
+		return a
+	}
+	return b
+}

internal/parser/vendors/nvidia/parser.go

@@ -14,7 +14,7 @@ import (
 
 // parserVersion - version of this parser module
 // IMPORTANT: Increment this version when making changes to parser logic!
-const parserVersion = "1.1.0"
+const parserVersion = "1.2.0"
 
 func init() {
 	parser.Register(&Parser{})
@@ -124,6 +124,15 @@ func (p *Parser) Parse(files []parser.ExtractedFile) (*models.AnalysisResult, er
 		}
 	}
 
+	// Parse inventory/output.log (contains GPU serial numbers from lspci)
+	inventoryLogFile := findInventoryOutputLog(files)
+	if inventoryLogFile != nil {
+		if err := ParseInventoryLog(inventoryLogFile.Content, result); err != nil {
+			// Log error but continue parsing other files
+			_ = err // Ignore error for now
+		}
+	}
+
 	// Parse summary.json (test results summary)
 	if f := parser.FindFileByName(files, "summary.json"); f != nil {
 		events := ParseSummaryJSON(f.Content)

internal/parser/vendors/nvidia/parser_test.go

@@ -0,0 +1,145 @@
+package nvidia
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"git.mchus.pro/mchus/logpile/internal/parser"
+)
+
+func TestNVIDIAParser_RealArchive(t *testing.T) {
+	// Test with the real archive that was reported as problematic
+	archivePath := filepath.Join("../../../../example", "A514359X5A09844_logs-20260115-151707.tar")
+
+	// Check if file exists
+	if _, err := os.Stat(archivePath); os.IsNotExist(err) {
+		t.Skip("Test archive not found, skipping test")
+	}
+
+	// Extract files from archive
+	files, err := parser.ExtractArchive(archivePath)
+	if err != nil {
+		t.Fatalf("Failed to extract archive: %v", err)
+	}
+
+	// Check if inventory/output.log exists
+	hasInventoryLog := false
+	for _, f := range files {
+		if filepath.Base(f.Path) == "output.log" {
+			t.Logf("Found file: %s", f.Path)
+		}
+		if f.Path == "./inventory/output.log" || f.Path == "inventory/output.log" {
+			hasInventoryLog = true
+			t.Logf("Found inventory/output.log with %d bytes", len(f.Content))
+		}
+	}
+	if !hasInventoryLog {
+		t.Error("inventory/output.log not found in extracted files")
+	}
+
+	// Create parser and parse
+	p := &Parser{}
+	result, err := p.Parse(files)
+	if err != nil {
+		t.Fatalf("Failed to parse archive: %v", err)
+	}
+
+	// Verify basic system info
+	if result.Hardware.BoardInfo.Manufacturer == "" {
+		t.Error("Expected Manufacturer to be set")
+	}
+	if result.Hardware.BoardInfo.ProductName == "" {
+		t.Error("Expected ProductName to be set")
+	}
+	if result.Hardware.BoardInfo.SerialNumber == "" {
+		t.Error("Expected SerialNumber to be set")
+	}
+
+	t.Logf("System Info:")
+	t.Logf("  Manufacturer: %s", result.Hardware.BoardInfo.Manufacturer)
+	t.Logf("  Product: %s", result.Hardware.BoardInfo.ProductName)
+	t.Logf("  Serial: %s", result.Hardware.BoardInfo.SerialNumber)
+
+	// Verify GPUs were found
+	if len(result.Hardware.GPUs) == 0 {
+		t.Error("Expected to find GPUs")
+	}
+
+	t.Logf("\nFound %d GPUs:", len(result.Hardware.GPUs))
+
+	gpusWithSerials := 0
+	for _, gpu := range result.Hardware.GPUs {
+		t.Logf("  %s: %s (Firmware: %s, Serial: %s, BDF: %s)",
+			gpu.Slot, gpu.Model, gpu.Firmware, gpu.SerialNumber, gpu.BDF)
+
+		if gpu.SerialNumber != "" {
+			gpusWithSerials++
+		}
+	}
+
+	// Verify that GPU serial numbers were extracted
+	if gpusWithSerials == 0 {
+		t.Error("Expected at least some GPUs to have serial numbers")
+	}
+
+	t.Logf("\nGPUs with serial numbers: %d/%d", gpusWithSerials, len(result.Hardware.GPUs))
+
+	// Check events for SXM2 failures
+	t.Logf("\nTotal events: %d", len(result.Events))
+
+	// Look for the specific serial or SXM2
+	sxm2Events := 0
+	for _, event := range result.Events {
+		desc := event.Description + " " + event.RawData + " " + event.EventType
+		if contains(desc, "SXM2") || contains(desc, "1653925025827") {
+			t.Logf("  SXM2 Event: [%s] %s (Severity: %s)", event.EventType, event.Description, event.Severity)
+			sxm2Events++
+		}
+	}
+
+	if sxm2Events == 0 {
+		t.Error("Expected to find events for SXM2 (faulty GPU 1653925025827)")
+	}
+	t.Logf("\nSXM2 failure events: %d", sxm2Events)
+}
+
+func contains(s, substr string) bool {
+	return len(s) >= len(substr) && (s == substr || len(s) > len(substr) &&
+		(s[:len(substr)] == substr || s[len(s)-len(substr):] == substr ||
+		 findSubstring(s, substr)))
+}
+
+func findSubstring(s, substr string) bool {
+	for i := 0; i <= len(s)-len(substr); i++ {
+		if s[i:i+len(substr)] == substr {
+			return true
+		}
+	}
+	return false
+}
+
+func TestFormatGPUSerial(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected string
+	}{
+		{
+			input:    "14-17-dc-65-77-2d-b0-48",
+			expected: "48:B0:2D:77:65:DC:17:14",
+		},
+		{
+			input:    "f2-fd-85-e0-2f-2d-b0-48",
+			expected: "48:B0:2D:2F:E0:85:FD:F2",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.input, func(t *testing.T) {
+			result := formatGPUSerial(tt.input)
+			if result != tt.expected {
+				t.Errorf("formatGPUSerial(%s) = %s, want %s", tt.input, result, tt.expected)
+			}
+		})
+	}
+}