fix(pcie): stop misclassifying same-vendor non-GPU devices as GPUs

matchesGPUVendor (sat_overlay.go) matched any PCIe DeviceClass containing
the substring "Controller" instead of the actual GPU classes, so a
same-vendor non-GPU device (e.g. an NVIDIA-branded NIC/storage controller)
with a degraded PCIe link surfaced as a pcie:gpu:<vendor> alarm on the
Hardware Summary/webui.

isGPUDevice (app_format.go) had the same bug via a different path: an
"any NVIDIA-vendor device is a GPU" fallback that could inflate the
"GPU: N x <model>" audit summary line with non-GPU companion devices
(NVSwitch/NVLink bridges, etc).

Both now defer to collector.IsGPUClass (exported from the previously
unexported isGPUClass), the canonical exact-match classifier already used
by amdgpu.go — removing the duplicated/incorrect class checks instead of
reimplementing them. Also dedup two more locally-duplicated classifiers
within webui (which doesn't import collector by convention): NIC-class
matching (pages.go/page_topo.go) and GPU-class matching in
page_validate.go now share one copy per package instead of being
reimplemented per file.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

audit/internal/app/app_format.go

@@ -134,14 +134,11 @@ func isGPUDevice(dev schema.HardwarePCIeDevice) bool {
 	if dev.VendorID != nil && *dev.VendorID == collector.AspeedVendorID {
 		return false
 	}
-	class := trimPtr(dev.DeviceClass)
+	// AMD Instinct / Radeon and NVIDIA compute GPUs always carry VideoController,
-	// AMD Instinct / Radeon compute GPUs always carry ProcessingAccelerator or DisplayController.
+	// DisplayController, or ProcessingAccelerator. Do NOT match vendor ID alone —
-	// Do NOT match AMD vendor alone — CPU chipset PCIe devices share that vendor ID.
+	// same-vendor companion devices (NVSwitch/NVLink bridges, CPU chipset PCIe
-	if class == "VideoController" || class == "DisplayController" || class == "ProcessingAccelerator" {
+	// functions, ...) share the GPU's vendor ID without being a GPU die.
-		return true
+	return collector.IsGPUClass(trimPtr(dev.DeviceClass))
-	}
-	// NVIDIA devices sometimes expose class values outside the standard GPU set.
-	return dev.VendorID != nil && *dev.VendorID == collector.NvidiaVendorID
 }
 
 func formatSystemLine(board schema.HardwareBoard) string {

audit/internal/app/app_format_test.go

@@ -0,0 +1,32 @@
+package app
+
+import (
+	"testing"
+
+	"bee/audit/internal/schema"
+)
+
+// TestFormatGPULineIgnoresNonGPUSameVendorDevice guards the bug where
+// isGPUDevice treated any NVIDIA-vendor PCIe device as a GPU regardless of
+// class (a fallback for "NVIDIA devices sometimes expose class values
+// outside the standard GPU set"). That fallback misclassified same-vendor
+// companion devices — e.g. NVSwitch/NVLink bridges, or an NVIDIA-branded NIC
+// — as compute GPUs, inflating the "GPU: N x <model>" audit summary line.
+func TestFormatGPULineIgnoresNonGPUSameVendorDevice(t *testing.T) {
+	vendor := 0x10de // collector.NvidiaVendorID
+	bridgeClass := "Bridge"
+	bridgeModel := "NVSwitch"
+	gpuClass := "VideoController"
+	gpuModel := "H100"
+
+	devices := []schema.HardwarePCIeDevice{
+		{DeviceClass: &bridgeClass, VendorID: &vendor, Model: &bridgeModel},
+		{DeviceClass: &gpuClass, VendorID: &vendor, Model: &gpuModel},
+	}
+
+	got := formatGPULine(devices)
+	want := "GPU: 1 x H100"
+	if got != want {
+		t.Fatalf("formatGPULine() = %q, want %q", got, want)
+	}
+}

audit/internal/app/app_test.go

@@ -391,6 +391,38 @@ func TestWritePCIeGPUStatusesToDBSurfacesLinkSpeedDegradation(t *testing.T) {
 	}
 }
 
+// TestWritePCIeGPUStatusesToDBIgnoresNonGPUSameVendorDevice guards the bug
+// where a PCIe link-speed degradation on a non-GPU device that merely shares
+// the GPU's PCI vendor ID (e.g. an NVIDIA-branded NIC/storage/NVLink-bridge
+// companion device) was misclassified as a GPU fault and surfaced as a
+// pcie:gpu:<vendor> alarm, because matchesGPUVendor matched any class
+// containing the substring "Controller" instead of the actual GPU classes.
+func TestWritePCIeGPUStatusesToDBIgnoresNonGPUSameVendorDevice(t *testing.T) {
+	db, err := OpenComponentStatusDB(filepath.Join(t.TempDir(), "component-status.json"))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	class := "NetworkController"
+	vendor := 0x10de // collector.NvidiaVendorID
+	status := "Warning"
+	desc := "PCIe link speed degraded: running at Gen1, capable of Gen5"
+	devices := []schema.HardwarePCIeDevice{
+		{
+			HardwareComponentStatus: schema.HardwareComponentStatus{Status: &status, ErrorDescription: &desc},
+			DeviceClass:             &class,
+			VendorID:                &vendor,
+			BDF:                     strPtr("0000:02:00.0"),
+		},
+	}
+
+	writePCIeGPUStatusesToDB(db, devices)
+
+	if _, ok := db.Get("pcie:gpu:nvidia"); ok {
+		t.Fatal("expected non-GPU same-vendor device to not write pcie:gpu:nvidia")
+	}
+}
+
 func TestRunNCCLTestsPassesSelectedGPUs(t *testing.T) {
 	t.Parallel()
 

audit/internal/app/sat_overlay.go

@@ -340,9 +340,7 @@ func matchesGPUVendor(dev schema.HardwarePCIeDevice, vendor string) bool {
 		return false
 	}
 	class := strings.TrimSpace(*dev.DeviceClass)
-	isGPUClass := strings.Contains(class, "Controller") || strings.Contains(class, "Accelerator") ||
+	if !collector.IsGPUClass(class) {
-		strings.Contains(class, "Display") || strings.Contains(class, "Video")
-	if !isGPUClass {
 		return false
 	}
 	switch vendor {

audit/internal/collector/amdgpu.go

@@ -87,7 +87,7 @@ func isAMDGPUDevice(dev schema.HardwarePCIeDevice) bool {
 	if dev.DeviceClass == nil {
 		return false
 	}
-	return dev.VendorID != nil && *dev.VendorID == AMDVendorID && isGPUClass(strings.TrimSpace(*dev.DeviceClass))
+	return dev.VendorID != nil && *dev.VendorID == AMDVendorID && IsGPUClass(strings.TrimSpace(*dev.DeviceClass))
 }
 
 func queryAMDGPUs() (map[string]amdGPUInfo, error) {

audit/internal/collector/contract.go

@@ -48,7 +48,11 @@ func isNICClass(class string) bool {
 	}
 }
 
-func isGPUClass(class string) bool {
+// IsGPUClass reports whether a normalized PCIe device class (see
+// mapPCIeDeviceClass) identifies a GPU/accelerator die itself, as opposed to
+// a same-vendor companion device (NIC, storage controller, NVLink bridge,
+// audio/USB function, ...) that happens to share the GPU's PCI vendor ID.
+func IsGPUClass(class string) bool {
 	switch strings.TrimSpace(class) {
 	case "VideoController", "DisplayController", "ProcessingAccelerator":
 		return true

audit/internal/collector/pcie_identity.go

@@ -51,7 +51,7 @@ func shouldProbePCIeSerial(dev schema.HardwarePCIeDevice) bool {
 		return false
 	}
 	class := strings.TrimSpace(*dev.DeviceClass)
-	return isNICClass(class) || isGPUClass(class)
+	return isNICClass(class) || IsGPUClass(class)
 }
 
 func queryPCIDeviceSerial(bdf string) string {

audit/internal/webui/page_topo.go

@@ -54,15 +54,12 @@ func topoCard(title, body string) string {
 // instead of importing the package for one classifier).
 // ---------------------------------------------------------------------------
 
-// isNICDeviceClassDev mirrors the classification logic in hwDescribeNIC
+// isNICDeviceClassDev applies isNICDeviceClass (pages.go) to a single device,
-// (pages.go), applied to a single device instead of aggregated counts.
+// with a MAC-address fallback for devices lspci doesn't classify as NIC.
 func isNICDeviceClassDev(dev schema.HardwarePCIeDevice) bool {
-	if dev.DeviceClass != nil {
+	if dev.DeviceClass != nil && isNICDeviceClass(*dev.DeviceClass) {
-		c := strings.ToLower(strings.TrimSpace(*dev.DeviceClass))
-		if c == "ethernetcontroller" || c == "networkcontroller" || strings.Contains(c, "fibrechannel") {
 		return true
 	}
-	}
 	return len(dev.MacAddresses) > 0
 }
 

audit/internal/webui/page_validate.go

@@ -627,8 +627,7 @@ func validateIsVendorGPU(dev schema.HardwarePCIeDevice, vendor string) bool {
 	if dev.VendorID != nil && *dev.VendorID == pciVendorAspeed {
 		return false
 	}
-	class := strings.ToLower(validateTrimPtr(dev.DeviceClass))
+	isGPUClass := isGPUDeviceClass(validateTrimPtr(dev.DeviceClass))
-	isGPUClass := class == "videocontroller" || class == "processingaccelerator" || class == "displaycontroller"
 	switch vendor {
 	case "nvidia":
 		return isGPUClass && dev.VendorID != nil && *dev.VendorID == pciVendorNvidia

audit/internal/webui/pages.go

@@ -560,16 +560,21 @@ func hwDescribePSU(hw schema.HardwareSnapshot) string {
 	return fmt.Sprintf("%d× PSU", n)
 }
 
+// isNICDeviceClass reports whether a normalized PCIe DeviceClass string
+// (see collector.mapPCIeDeviceClass) identifies a NIC/HBA-class device.
+// webui does not import collector (see the "Classification helpers" note
+// in page_topo.go), so this mirrors collector.isNICClass locally.
+func isNICDeviceClass(class string) bool {
+	c := strings.ToLower(strings.TrimSpace(class))
+	return c == "ethernetcontroller" || c == "networkcontroller" || strings.Contains(c, "fibrechannel")
+}
+
 // hwDescribeNIC returns a summary like "2× Mellanox ConnectX-6".
 func hwDescribeNIC(hw schema.HardwareSnapshot) string {
 	counts := map[string]int{}
 	order := []string{}
 	for _, dev := range hw.PCIeDevices {
-		isNIC := false
+		isNIC := dev.DeviceClass != nil && isNICDeviceClass(*dev.DeviceClass)
-		if dev.DeviceClass != nil {
-			c := strings.ToLower(strings.TrimSpace(*dev.DeviceClass))
-			isNIC = c == "ethernetcontroller" || c == "networkcontroller" || strings.Contains(c, "fibrechannel")
-		}
 		if !isNIC && len(dev.MacAddresses) == 0 {
 			continue
 		}