INC-2026-05-23-002: Investigation

Root Cause

The laptop display (eDP-1) becomes unrecoverable after i915 Panel Self Refresh (PSR) enters a hung state on kernel 7.0.9. This is an i915 driver regression specific to Arrow Lake-S Intel Graphics, not NVIDIA.

What is PSR?

Panel Self Refresh is an eDP power-saving feature where the GPU writes a frame to the panel’s internal framebuffer, then the panel refreshes itself without GPU involvement. When the GPU needs to update the display, it signals the panel to exit PSR. On kernel 7.0.9, the PSR exit signal times out, leaving the display pipe locked.

Display Topology

DRM Device GPU Connected Displays

card0

NVIDIA GeForce RTX 5090 (nvidia-open)

None — DP-5, DP-6, HDMI-A-1 all disabled

card1

Intel Arrow Lake-S (i915)

eDP-1 (laptop panel), DP-1 (external via USB-C/Thunderbolt)

eDP-1 is driven entirely by the Intel iGPU. The NVIDIA Cannot find any crtc or sizes error logged at boot is normal for hybrid GPU laptops where no displays are wired to the discrete GPU.

Failure Mechanism

  1. i915 enables PSR for eDP-1 (default behavior for eDP panels)

  2. During idle, PSR enters self-refresh mode

  3. On resume, i915 attempts to signal PSR exit

  4. PSR exit times out: Timed out waiting PSR idle state

  5. Display pipe cannot be cycled: pipe_off wait timed out

  6. Kernel WARNING fires in intel_disable_transcoder

  7. eDP-1 is locked — no method (DPMS, monitor disable, brightness) can recover it

  8. External monitors on separate display pipes are unaffected

Misdiagnosis Sequence

Attempt Hypothesis Result

1

NVIDIA DRM CRTC failure — Cannot find any crtc or sizes

Wrong — eDP-1 is on Intel iGPU, not NVIDIA. Error is normal for hybrid GPU.

2

DPMS wake failure — replaced with monitor disable/enable

Failed — same i915 DRM path, same PSR hang

3

DRM re-initialization failure — replaced with brightnessctl (no DRM interaction)

Failed — PSR hangs the display pipe independently of any userspace mechanism

4

i915 PSR hang — disabled PSR via i915.enable_psr=0

Correct root cause — kernel parameter applied, pending reboot confirmation

Packages Upgraded (2026-05-22)

Package From To

linux

6.19.14.arch1-1

7.0.9.arch1-1

linux-headers

6.19.14.arch1-1

7.0.9.arch1-1

nvidia-open

595.58.03-6

595.71.05-10

nvidia-utils

595.58.03-2

595.71.05-2

mesa

26.0.5-1

26.1.1-2

libdrm

2.4.131-1

2.4.133-1

hyprland

0.54.3-3

0.55.2-1

Evidence

i915 PSR timeout (definitive — this is the root cause)

May 23 14:11:01 modestus-p16g kernel: i915 0000:00:02.0: [drm] *ERROR* Timed out waiting PSR idle state
May 23 14:11:55 modestus-p16g kernel: i915 0000:00:02.0: [drm] pipe_off wait timed out
May 23 14:11:55 modestus-p16g kernel: WARNING: drivers/gpu/drm/i915/display/intel_display.c:363
    at intel_disable_transcoder+0x2da/0x330 [i915], CPU#14: Hyprland/13909

i915 kernel stack trace (display pipe failure)

RIP: 0010:intel_disable_transcoder+0x2e0/0x330 [i915]
  intel_ddi_post_disable+0x38d/0x6f0 [i915]
  intel_encoders_post_disable+0x72/0x90 [i915]
  hsw_crtc_disable+0x55/0x100 [i915]
  intel_old_crtc_state_disables+0xdf/0x1e0 [i915]
  intel_atomic_commit_tail+0x166c/0x18a0 [i915]
  intel_atomic_commit+0x23d/0x280 [i915]
  drm_atomic_commit+0xb1/0xe0
  drm_mode_atomic_ioctl+0xb65/0xdb0

DRM topology (eDP-1 on Intel iGPU)

$ ls -la /sys/class/drm/card1-eDP-1
lrwxrwxrwx ... card1-eDP-1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-eDP-1

$ lspci -k -s 00:02.0
00:02.0 VGA compatible controller: Intel Corporation Arrow Lake-S [Intel Graphics] (rev 06)
    Kernel driver in use: i915
    Kernel modules: i915, xe

NVIDIA error (red herring — normal for hybrid GPU)

May 23 09:27:04 modestus-p16g kernel: nvidia 0000:01:00.0: [drm] Cannot find any crtc or sizes

This is expected: no monitors are wired to the NVIDIA GPU on this laptop.

Open Questions

  • Does kernel 6.18.32 (linux-lts) exhibit the same i915 PSR hang?

  • Will a future kernel update fix PSR for Arrow Lake-S, allowing i915.enable_psr=0 to be removed?

  • Is the xe driver (also available for this GPU) affected by the same PSR regression?