From 481b67fb69675099f9fc5c33210e93043244f27d Mon Sep 17 00:00:00 2001
From: josh <josh@thewrightserver.net>
Date: Sun, 19 Apr 2026 21:56:18 -0400
Subject: [PATCH] feat(firmware): install probe tools in live image + surface
 nic/hba gaps

mkosi.conf: add ipmitool, ethtool, nvme-cli so the Firmware stage
can actually read BMC revisions, NIC firmware versions, and fall
back to nvme-cli when sysfs firmware_rev is missing.

firmware.go: probeNICFirmware and probeHBAFirmware now return
(snapshots, warning) so a missing ethtool/lspci surfaces in the
stage log the same way probeBIOS/probeBMC already do. Before, a
host without ethtool silently reported "bios=1 nvme_fw=1
microcode=1" with no hint that nic coverage was dropped.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---
 agent/probes/firmware.go | 28 ++++++++++++++++++----------
 live-image/mkosi.conf    |  6 ++++++
 2 files changed, 24 insertions(+), 10 deletions(-)

diff --git a/agent/probes/firmware.go b/agent/probes/firmware.go
index db4c37e..7ccf52a 100644
--- a/agent/probes/firmware.go
+++ b/agent/probes/firmware.go
@@ -47,9 +47,17 @@ func Firmware(ctx context.Context) ([]FirmwareSnapshot, []string) {
 	} else if warn != "" {
 		warnings = append(warnings, warn)
 	}
-	out = append(out, probeNICFirmware(ctx)...)
+	nicSnaps, nicWarn := probeNICFirmware(ctx)
+	out = append(out, nicSnaps...)
+	if nicWarn != "" {
+		warnings = append(warnings, nicWarn)
+	}
 	out = append(out, probeNVMeFirmware(ctx)...)
-	out = append(out, probeHBAFirmware(ctx)...)
+	hbaSnaps, hbaWarn := probeHBAFirmware(ctx)
+	out = append(out, hbaSnaps...)
+	if hbaWarn != "" {
+		warnings = append(warnings, hbaWarn)
+	}
 	if snap := probeMicrocode(); snap != nil {
 		out = append(out, *snap)
 	}
@@ -214,13 +222,13 @@ func parseIpmitoolMCInfo(r io.Reader) *FirmwareSnapshot {
 // `ethtool -i <iface>` on each real NIC (skip lo, bridges, virtuals).
 // One snapshot per interface so a mismatched port lights up in the diff
 // without silencing sibling ports.
-func probeNICFirmware(ctx context.Context) []FirmwareSnapshot {
+func probeNICFirmware(ctx context.Context) ([]FirmwareSnapshot, string) {
 	if _, err := exec.LookPath("ethtool"); err != nil {
-		return nil
+		return nil, "nic: ethtool not installed"
 	}
 	ifaces, err := os.ReadDir("/sys/class/net")
 	if err != nil {
-		return nil
+		return nil, ""
 	}
 	var out []FirmwareSnapshot
 	for _, entry := range ifaces {
@@ -237,7 +245,7 @@ func probeNICFirmware(ctx context.Context) []FirmwareSnapshot {
 			out = append(out, *snap)
 		}
 	}
-	return out
+	return out, ""
 }
 
 // parseEthtoolI extracts driver/firmware-version from `ethtool -i`
@@ -353,15 +361,15 @@ var lspciClassHBA = regexp.MustCompile(`(?i)(serial attached scsi|sas controller
 // "revision" on the device line. We capture what's printed and rely on
 // SpecValidate to diff — this keeps us off tool-specific CLIs (storcli,
 // mpt-status) that aren't always installed.
-func probeHBAFirmware(ctx context.Context) []FirmwareSnapshot {
+func probeHBAFirmware(ctx context.Context) ([]FirmwareSnapshot, string) {
 	if _, err := exec.LookPath("lspci"); err != nil {
-		return nil
+		return nil, "hba: lspci not installed"
 	}
 	out, err := runCmd(ctx, "lspci", "-Dvvnn")
 	if err != nil {
-		return nil
+		return nil, fmt.Sprintf("hba: lspci failed: %v", trimErr(err, out))
 	}
-	return parseLspciHBA(strings.NewReader(out))
+	return parseLspciHBA(strings.NewReader(out)), ""
 }
 
 // parseLspciHBA walks `lspci -Dvvnn` stanzas and picks SAS/RAID
diff --git a/live-image/mkosi.conf b/live-image/mkosi.conf
index ba96c44..cda66d9 100644
--- a/live-image/mkosi.conf
+++ b/live-image/mkosi.conf
@@ -57,6 +57,12 @@ Packages=
     lm-sensors
     e2fsprogs
     util-linux
+    # Firmware probe tooling. Without these, the Firmware stage silently
+    # skips whole components (ethtool → nic, nvme-cli → nvme fallback) or
+    # emits a cosmetic "not installed" warning (ipmitool → bmc).
+    ipmitool
+    ethtool
+    nvme-cli
     # Firmware. firmware-linux-nonfree on bookworm is a thin metapackage
     # that does NOT pull i915 GuC/HuC — those live in firmware-misc-nonfree.
     # Enumerate explicitly so the blob for whatever hardware we boot on