Compare commits

...

5 Commits
v5.8 ... v5.9

Author SHA1 Message Date
7a21c370e4 Handle NVIDIA GSP firmware init hang with timeout fallback
- bee-nvidia-load: run insmod in background, poll /proc/devices for
  nvidiactl; if GSP init doesn't complete in 90s, kill insmod and retry
  with NVreg_EnableGpuFirmware=0. Handles EBUSY case with clear error.
- Write /run/bee-nvidia-mode (gsp-on/gsp-off/gsp-stuck) for audit layer
- Show GSP mode badge in sidebar: yellow for gsp-off, red for gsp-stuck
- Report NvidiaGSPMode in RuntimeHealth with issue entries
- Simplify GRUB menu: default (KMS+GSP), advanced submenu (GSP=off,
  nomodeset, fail-safe), remove load-to-RAM entry
- Add pcmanfm, ristretto, mupdf, mousepad to desktop packages

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 21:00:43 +03:00
a493e3ab5b Fix service control buttons: sudo, real error output, UX feedback
- services.go: use sudo systemctl so bee user can control system services
- api.go: always return 200 with output field even on error, so the
  frontend shows the actual systemctl message instead of "exit status 1"
- pages.go: button shows "..." while pending then restores label;
  output panel is full-width under the table with ✓/✗ status indicator;
  output auto-scrolls to bottom

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 20:25:41 +03:00
19b4803ec7 Pass exact cycle duration to GPU stress instead of 86400s sentinel
bee-gpu-burn now receives --seconds <LoadSec> so it exits naturally
when the cycle ends, rather than relying solely on context cancellation
to kill it. Process group kill (Setpgid+Cancel) is kept as a safety net
for early cancellation (user stop, context timeout). Same fix for AMD
RVS which now gets duration_ms = LoadSec * 1000.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 20:22:43 +03:00
1bdfb1e9ca Fix nvidia-targeted-stress failing with DCGM_ST_IN_USE (-34)
nvvs (DCGM validation suite) survives when dcgmi is killed mid-run,
leaving the GPU occupied. The next dcgmi diag invocation then fails
with "affected resource is in use".

Two-part fix:
- Add nvvs and dcgmi to KillTestWorkers patterns so they are cleaned
  up by the global cancel handler
- Call KillTestWorkers at the start of RunNvidiaTargetedStressValidatePack
  to clear any stale processes before dcgmi diag runs

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 20:21:36 +03:00
c5d6b30177 Fix platform thermal cycling leaving GPU load running after test ends
bee-gpu-burn is a shell script that spawns bee-gpu-burn-worker children.
exec.CommandContext default cancel only kills the shell parent; the worker
processes survive and keep loading the GPU indefinitely.

Fix: set Setpgid=true and a custom Cancel that sends SIGKILL to the
entire process group (-pid), same pattern already used in runSATCommandCtx.
Applied to Nvidia, AMD, and CPU stress commands for consistency.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-05 20:19:20 +03:00
11 changed files with 227 additions and 49 deletions

View File

@@ -15,6 +15,10 @@ var workerPatterns = []string{
"stress-ng",
"stressapptest",
"memtester",
// DCGM diagnostic workers — nvvs is spawned by dcgmi diag and survives
// if dcgmi is killed mid-run, leaving the GPU occupied (DCGM_ST_IN_USE).
"nvvs",
"dcgmi",
}
// KilledProcess describes a process that was sent SIGKILL.

View File

@@ -110,7 +110,7 @@ func (s *System) RunPlatformStress(
wg.Add(1)
go func() {
defer wg.Done()
gpuCmd := buildGPUStressCmd(loadCtx, vendor)
gpuCmd := buildGPUStressCmd(loadCtx, vendor, cycle.LoadSec)
if gpuCmd == nil {
return
}
@@ -392,6 +392,13 @@ func buildCPUStressCmd(ctx context.Context) (*exec.Cmd, error) {
cmdArgs = append(cmdArgs, "-M", strconv.Itoa(mb))
}
cmd := exec.CommandContext(ctx, path, cmdArgs...)
cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
cmd.Cancel = func() error {
if cmd.Process != nil {
_ = syscall.Kill(-cmd.Process.Pid, syscall.SIGKILL)
}
return nil
}
cmd.Stdout = nil
cmd.Stderr = nil
if err := startLowPriorityCmd(cmd, 15); err != nil {
@@ -402,28 +409,28 @@ func buildCPUStressCmd(ctx context.Context) (*exec.Cmd, error) {
// buildGPUStressCmd creates a GPU stress command appropriate for the detected vendor.
// Returns nil if no GPU stress tool is available (CPU-only cycling still useful).
func buildGPUStressCmd(ctx context.Context, vendor string) *exec.Cmd {
func buildGPUStressCmd(ctx context.Context, vendor string, durSec int) *exec.Cmd {
switch strings.ToLower(vendor) {
case "amd":
return buildAMDGPUStressCmd(ctx)
return buildAMDGPUStressCmd(ctx, durSec)
case "nvidia":
return buildNvidiaGPUStressCmd(ctx)
return buildNvidiaGPUStressCmd(ctx, durSec)
}
return nil
}
func buildAMDGPUStressCmd(ctx context.Context) *exec.Cmd {
func buildAMDGPUStressCmd(ctx context.Context, durSec int) *exec.Cmd {
rvsArgs, err := resolveRVSCommand()
if err != nil {
return nil
}
rvsPath := rvsArgs[0]
cfg := `actions:
cfg := fmt.Sprintf(`actions:
- name: gst_platform
device: all
module: gst
parallel: true
duration: 86400000
duration: %d`, durSec*1000) + `
copy_matrix: false
target_stress: 90
matrix_size_a: 8640
@@ -433,13 +440,20 @@ func buildAMDGPUStressCmd(ctx context.Context) *exec.Cmd {
cfgFile := "/tmp/bee-platform-gst.conf"
_ = os.WriteFile(cfgFile, []byte(cfg), 0644)
cmd := exec.CommandContext(ctx, rvsPath, "-c", cfgFile)
cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
cmd.Cancel = func() error {
if cmd.Process != nil {
_ = syscall.Kill(-cmd.Process.Pid, syscall.SIGKILL)
}
return nil
}
cmd.Stdout = nil
cmd.Stderr = nil
_ = startLowPriorityCmd(cmd, 10)
return cmd
}
func buildNvidiaGPUStressCmd(ctx context.Context) *exec.Cmd {
func buildNvidiaGPUStressCmd(ctx context.Context, durSec int) *exec.Cmd {
path, err := satLookPath("bee-gpu-burn")
if err != nil {
path, err = satLookPath("bee-gpu-stress")
@@ -447,7 +461,17 @@ func buildNvidiaGPUStressCmd(ctx context.Context) *exec.Cmd {
if err != nil {
return nil
}
cmd := exec.CommandContext(ctx, path, "--seconds", "86400")
// Pass exact duration so bee-gpu-burn exits on its own when the cycle ends.
// Process group kill via Setpgid+Cancel is kept as a safety net for cases
// where the context is cancelled early (user stop, parent timeout).
cmd := exec.CommandContext(ctx, path, "--seconds", strconv.Itoa(durSec))
cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
cmd.Cancel = func() error {
if cmd.Process != nil {
_ = syscall.Kill(-cmd.Process.Pid, syscall.SIGKILL)
}
return nil
}
cmd.Stdout = nil
cmd.Stderr = nil
_ = startLowPriorityCmd(cmd, 10)

View File

@@ -173,6 +173,22 @@ func (s *System) collectGPURuntimeHealth(vendor string, health *schema.RuntimeHe
switch vendor {
case "nvidia":
if raw, err := os.ReadFile("/run/bee-nvidia-mode"); err == nil {
health.NvidiaGSPMode = strings.TrimSpace(string(raw))
if health.NvidiaGSPMode == "gsp-stuck" {
health.Issues = append(health.Issues, schema.RuntimeIssue{
Code: "nvidia_gsp_stuck",
Severity: "critical",
Description: "NVIDIA GSP firmware init timed out and the kernel module is stuck. Reboot and select 'GSP=off' in the boot menu.",
})
} else if health.NvidiaGSPMode == "gsp-off" {
health.Issues = append(health.Issues, schema.RuntimeIssue{
Code: "nvidia_gsp_disabled",
Severity: "warning",
Description: "NVIDIA GSP firmware disabled (fallback). Power management runs via CPU path — power draw readings may differ from reference hardware.",
})
}
}
health.DriverReady = strings.Contains(lsmodText, "nvidia ")
if !health.DriverReady {
health.Issues = append(health.Issues, schema.RuntimeIssue{

View File

@@ -382,6 +382,13 @@ func (s *System) RunNvidiaTargetedStressValidatePack(ctx context.Context, baseDi
if err != nil {
return "", err
}
// Kill any lingering nvvs/dcgmi processes from a previous interrupted run
// before starting — otherwise dcgmi diag fails with DCGM_ST_IN_USE (-34).
if killed := KillTestWorkers(); len(killed) > 0 && logFunc != nil {
for _, p := range killed {
logFunc(fmt.Sprintf("pre-flight: killed stale worker pid=%d name=%s", p.PID, p.Name))
}
}
return runAcceptancePackCtx(ctx, baseDir, "gpu-nvidia-targeted-stress", []satJob{
{name: "01-nvidia-smi-q.log", cmd: []string{"nvidia-smi", "-q"}},
{

View File

@@ -61,7 +61,9 @@ func (s *System) ServiceState(name string) string {
}
func (s *System) ServiceDo(name string, action ServiceAction) (string, error) {
raw, err := exec.Command("systemctl", string(action), name).CombinedOutput()
// bee-web runs as the bee user; sudo is required to control system services.
// /etc/sudoers.d/bee grants bee NOPASSWD:ALL.
raw, err := exec.Command("sudo", "systemctl", string(action), name).CombinedOutput()
return string(raw), err
}

View File

@@ -20,6 +20,7 @@ type RuntimeHealth struct {
ExportDir string `json:"export_dir,omitempty"`
DriverReady bool `json:"driver_ready,omitempty"`
CUDAReady bool `json:"cuda_ready,omitempty"`
NvidiaGSPMode string `json:"nvidia_gsp_mode,omitempty"` // "gsp-on", "gsp-off", "gsp-stuck"
NetworkStatus string `json:"network_status,omitempty"`
Issues []RuntimeIssue `json:"issues,omitempty"`
Tools []RuntimeToolStatus `json:"tools,omitempty"`

View File

@@ -383,11 +383,13 @@ func (h *handler) handleAPIServicesAction(w http.ResponseWriter, r *http.Request
return
}
result, err := h.opts.App.ServiceActionResult(req.Name, action)
status := "ok"
if err != nil {
writeError(w, http.StatusInternalServerError, err.Error())
return
status = "error"
}
writeJSON(w, map[string]string{"status": "ok", "output": result.Body})
// Always return 200 with output so the frontend can display the actual
// systemctl error message instead of a generic "exit status 1".
writeJSON(w, map[string]string{"status": status, "output": result.Body})
}
// ── Network ───────────────────────────────────────────────────────────────────

View File

@@ -33,6 +33,9 @@ a{color:var(--accent);text-decoration:none}
.sidebar-logo{padding:18px 16px 12px;font-size:18px;font-weight:700;color:#fff;letter-spacing:-.5px}
.sidebar-logo span{color:rgba(255,255,255,.5);font-weight:400;font-size:12px;display:block;margin-top:2px}
.sidebar-version{padding:0 16px 14px;font-size:11px;color:rgba(255,255,255,.45)}
.sidebar-badge{margin:0 12px 12px;padding:5px 8px;border-radius:4px;font-size:11px;font-weight:600;text-align:center}
.sidebar-badge-warn{background:#7a4f00;color:#f6c90e}
.sidebar-badge-crit{background:#5c1a1a;color:#ff6b6b}
.nav{flex:1}
.nav-item{display:block;padding:10px 16px;color:rgba(255,255,255,.7);font-size:13px;border-left:3px solid transparent;transition:all .15s}
.nav-item:hover{color:#fff;background:rgba(255,255,255,.08)}
@@ -107,6 +110,15 @@ func layoutNav(active string, buildLabel string) string {
buildLabel = "dev"
}
b.WriteString(`<div class="sidebar-version">Version ` + html.EscapeString(buildLabel) + `</div>`)
if raw, err := os.ReadFile("/run/bee-nvidia-mode"); err == nil {
gspMode := strings.TrimSpace(string(raw))
switch gspMode {
case "gsp-off":
b.WriteString(`<div class="sidebar-badge sidebar-badge-warn">NVIDIA GSP=off</div>`)
case "gsp-stuck":
b.WriteString(`<div class="sidebar-badge sidebar-badge-crit">NVIDIA GSP stuck — reboot</div>`)
}
}
b.WriteString(`<nav class="nav">`)
for _, item := range items {
cls := "nav-item"
@@ -2107,9 +2119,12 @@ func renderServicesInline() string {
return `<p style="font-size:13px;color:var(--muted);margin-bottom:10px">` + html.EscapeString(`bee-selfheal.timer is expected to be active; the oneshot bee-selfheal.service itself is not shown as a long-running service.`) + `</p>
<div style="display:flex;justify-content:flex-end;gap:8px;flex-wrap:wrap;margin-bottom:8px"><button class="btn btn-sm btn-secondary" onclick="restartGPUDrivers()">Restart GPU Drivers</button><button class="btn btn-sm btn-secondary" onclick="loadServices()">&#8635; Refresh</button></div>
<div id="svc-table"><p style="color:var(--muted);font-size:13px">Loading...</p></div>
<div id="svc-out" style="display:none;margin-top:8px" class="card">
<div class="card-head">Output</div>
<div class="card-body" style="padding:10px"><div id="svc-terminal" class="terminal" style="max-height:150px"></div></div>
<div id="svc-out" style="display:none;margin-top:12px">
<div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:4px">
<span id="svc-out-label" style="font-size:12px;font-weight:600;color:var(--muted)">Output</span>
<span id="svc-out-status" style="font-size:12px"></span>
</div>
<div id="svc-terminal" class="terminal" style="max-height:220px;width:100%;box-sizing:border-box"></div>
</div>
<script>
function loadServices() {
@@ -2125,9 +2140,9 @@ function loadServices() {
'<div id="'+id+'" style="display:none;margin-top:6px"><pre style="font-size:11px;white-space:pre-wrap;word-break:break-all;max-height:200px;overflow-y:auto;background:#1b1c1d;padding:8px;border-radius:4px;color:#b5cea8">'+body+'</pre></div>' +
'</td>' +
'<td style="white-space:nowrap">' +
'<button class="btn btn-sm btn-secondary" onclick="svcAction(\''+s.name+'\',\'start\')">Start</button> ' +
'<button class="btn btn-sm btn-secondary" onclick="svcAction(\''+s.name+'\',\'stop\')">Stop</button> ' +
'<button class="btn btn-sm btn-secondary" onclick="svcAction(\''+s.name+'\',\'restart\')">Restart</button>' +
'<button class="btn btn-sm btn-secondary" id="btn-'+s.name+'-start" onclick="svcAction(this,\''+s.name+'\',\'start\')">Start</button> ' +
'<button class="btn btn-sm btn-secondary" id="btn-'+s.name+'-stop" onclick="svcAction(this,\''+s.name+'\',\'stop\')">Stop</button> ' +
'<button class="btn btn-sm btn-secondary" id="btn-'+s.name+'-restart" onclick="svcAction(this,\''+s.name+'\',\'restart\')">Restart</button>' +
'</td></tr>';
}).join('');
document.getElementById('svc-table').innerHTML =
@@ -2138,16 +2153,45 @@ function toggleBody(id) {
const el = document.getElementById(id);
if (el) el.style.display = el.style.display==='none' ? 'block' : 'none';
}
function svcAction(name, action) {
function svcAction(btn, name, action) {
var label = btn.textContent;
btn.disabled = true;
btn.textContent = '...';
var out = document.getElementById('svc-out');
var term = document.getElementById('svc-terminal');
var statusEl = document.getElementById('svc-out-status');
var labelEl = document.getElementById('svc-out-label');
out.style.display = 'block';
labelEl.textContent = action + ' ' + name;
term.textContent = 'Running...';
statusEl.textContent = '';
statusEl.style.color = '';
fetch('/api/services/action',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({name,action})})
.then(r=>r.json()).then(d => {
document.getElementById('svc-out').style.display='block';
document.getElementById('svc-terminal').textContent = d.output || d.error || action+' '+name;
setTimeout(loadServices, 1000);
term.textContent = d.output || d.error || '(no output)';
term.scrollTop = term.scrollHeight;
if (d.status === 'ok') {
statusEl.textContent = '✓ done';
statusEl.style.color = 'var(--ok-fg, #2c662d)';
} else {
statusEl.textContent = '✗ failed';
statusEl.style.color = 'var(--crit-fg, #9f3a38)';
}
btn.textContent = label;
btn.disabled = false;
setTimeout(loadServices, 800);
}).catch(e => {
term.textContent = 'Request failed: ' + e;
statusEl.textContent = '✗ error';
statusEl.style.color = 'var(--crit-fg, #9f3a38)';
btn.textContent = label;
btn.disabled = false;
});
}
function restartGPUDrivers() {
svcAction('bee-nvidia', 'restart');
var btn = document.querySelector('[onclick*="restartGPUDrivers"]');
if (!btn) { svcAction({textContent:'',disabled:false}, 'bee-nvidia', 'restart'); return; }
svcAction(btn, 'bee-nvidia', 'restart');
}
loadServices();
</script>`

View File

@@ -11,33 +11,25 @@ echo " Hardware Audit LiveCD"
echo ""
menuentry "EASY-BEE" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ nomodeset bee.nvidia.mode=normal net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
linux @KERNEL_LIVE@ @APPEND_LIVE@ bee.nvidia.mode=normal net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
initrd @INITRD_LIVE@
}
menuentry "EASY-BEE (graphics/KMS)" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ bee.display=kms bee.nvidia.mode=normal net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
initrd @INITRD_LIVE@
}
submenu "EASY-BEE (advanced options) -->" {
menuentry "EASY-BEE — GSP=off" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ bee.nvidia.mode=gsp-off net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
initrd @INITRD_LIVE@
}
menuentry "EASY-BEE (load to RAM)" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ toram nomodeset bee.nvidia.mode=normal net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
initrd @INITRD_LIVE@
}
menuentry "EASY-BEE — nomodeset" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ nomodeset bee.nvidia.mode=normal net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
initrd @INITRD_LIVE@
}
menuentry "EASY-BEE (NVIDIA GSP=off)" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ nomodeset bee.nvidia.mode=gsp-off net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
initrd @INITRD_LIVE@
}
menuentry "EASY-BEE (graphics/KMS, GSP=off)" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ bee.display=kms bee.nvidia.mode=gsp-off net.ifnames=0 biosdevname=0 mitigations=off transparent_hugepage=always numa_balancing=disable nowatchdog nosoftlockup
initrd @INITRD_LIVE@
}
menuentry "EASY-BEE (fail-safe)" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ bee.nvidia.mode=gsp-off memtest noapic noapm nodma nomce nolapic nosmp vga=normal net.ifnames=0 biosdevname=0
initrd @INITRD_LIVE@
menuentry "EASY-BEE — fail-safe" {
linux @KERNEL_LIVE@ @APPEND_LIVE@ nomodeset bee.nvidia.mode=gsp-off noapic noapm nodma nomce nolapic nosmp vga=normal net.ifnames=0 biosdevname=0
initrd @INITRD_LIVE@
}
}
if [ "${grub_platform}" = "efi" ]; then

View File

@@ -65,6 +65,10 @@ python3-pil
xorg
xterm
chromium
mousepad
pcmanfm
ristretto
mupdf
xserver-xorg-video-fbdev
xserver-xorg-video-vesa
lightdm

View File

@@ -50,11 +50,93 @@ load_module() {
log "WARN: not found: $ko"
return 1
fi
if insmod "$ko" "$@"; then
if timeout 90 insmod "$ko" "$@"; then
log "loaded: $mod $*"
return 0
fi
log "WARN: failed to load: $mod"
log "WARN: failed to load: $mod (exit $?)"
dmesg | tail -n 10 | sed 's/^/ dmesg: /' || true
return 1
}
nvidia_is_functional() {
grep -q ' nvidiactl$' /proc/devices 2>/dev/null
}
load_module_with_gsp_fallback() {
ko="$NVIDIA_KO_DIR/nvidia.ko"
if [ ! -f "$ko" ]; then
log "ERROR: not found: $ko"
return 1
fi
# Run insmod in background — on some converted SXM→PCIe cards GSP enters an
# infinite crash/reload loop and insmod never returns. We check for successful
# initialization by polling /proc/devices for nvidiactl instead of waiting for
# insmod to exit.
log "loading nvidia (GSP enabled, timeout 90s)"
insmod "$ko" &
_insmod_pid=$!
_waited=0
while [ $_waited -lt 90 ]; do
if nvidia_is_functional; then
log "loaded: nvidia (GSP enabled, ${_waited}s)"
echo "gsp-on" > /run/bee-nvidia-mode
return 0
fi
# Check if insmod exited with an error before timeout
if ! kill -0 "$_insmod_pid" 2>/dev/null; then
wait "$_insmod_pid"
_rc=$?
if [ $_rc -ne 0 ]; then
log "nvidia load failed (exit $_rc)"
dmesg | tail -n 10 | sed 's/^/ dmesg: /' || true
return 1
fi
# insmod exited 0 but nvidiactl not yet in /proc/devices — give it a moment
sleep 2
if nvidia_is_functional; then
log "loaded: nvidia (GSP enabled, ${_waited}s)"
return 0
fi
log "insmod exited 0 but nvidiactl missing — treating as failure"
return 1
fi
sleep 1
_waited=$((_waited + 1))
done
# GSP init timed out — kill the hanging insmod and attempt gsp-off fallback
log "nvidia GSP init timed out after 90s"
kill "$_insmod_pid" 2>/dev/null || true
wait "$_insmod_pid" 2>/dev/null || true
# Attempt to unload the partially-initialized module
if ! rmmod nvidia 2>/dev/null; then
# Module is stuck in the kernel — cannot reload with different params.
# User must reboot and select bee.nvidia.mode=gsp-off at boot menu.
log "ERROR: rmmod nvidia failed (EBUSY) — module stuck in kernel"
log "ERROR: reboot and select 'EASY-BEE (advanced) -> GSP=off' in boot menu"
echo "gsp-stuck" > /run/bee-nvidia-mode
return 1
fi
sleep 2
log "retrying with NVreg_EnableGpuFirmware=0"
log "WARNING: GSP disabled — power management will run via CPU path, not GPU firmware"
if insmod "$ko" NVreg_EnableGpuFirmware=0; then
if nvidia_is_functional; then
log "loaded: nvidia (GSP disabled)"
echo "gsp-off" > /run/bee-nvidia-mode
return 0
fi
log "insmod gsp-off exited 0 but nvidiactl missing"
return 1
fi
log "nvidia load failed (GSP=off)"
dmesg | tail -n 10 | sed 's/^/ dmesg: /' || true
return 1
}
@@ -70,7 +152,7 @@ load_host_module() {
case "$nvidia_mode" in
normal|full)
if ! load_module nvidia; then
if ! load_module_with_gsp_fallback; then
exit 1
fi
# nvidia-modeset on some server kernels needs ACPI video helper symbols