Public bug reported: **Package:** linux-image-generic-hwe-26.04 7.0.0-28.28 **Regression:** yes, 7.0.0-27.27 -> 7.0.0-28.28 **Tags:** `regression-update kernel-bug i915 suspend-resume backlight resolute`
## Summary After upgrading to kernel 7.0.0-28, the internal display backlight does not come on when resuming from suspend. The machine is running after resume (it accepted a blind `sudo reboot` typed at the dark screen), but the backlight stays off. Increasing brightness with the brightness keys does not bring it on. The panel is still driving an image: screen contents are legible under a bright external light. The failure is therefore confined to the backlight, not a modeset or output failure. Booting 7.0.0-27 avoids the problem. Booting 7.0.0-28 with `i915.enable_dpcd_backlight=0` also avoids it. ## Hardware - **Machine:** Lenovo ThinkPad T480, machine type 20L6S0MD00 - **BIOS:** N24ET81W (1.56), 2025-09-06 - **GPU:** Intel Kaby Lake-R GT2 [UHD Graphics 620] `[8086:5917]` rev 07, subsystem `[17aa:225d]` (lspci reports "Lenovo ThinkPad T480 (20L5)") - **Panel:** LG Display LP140WF6-SPB7 on eDP-1 (model string read from EDID) - **OS:** Ubuntu 26.04 LTS (resolute) ## Steps to reproduce 1. Boot 7.0.0-28-generic with stock kernel parameters (`quiet splash`). 2. Suspend. The journal records `PM: suspend entry (deep)`. 3. Resume. **Expected:** display lights up. **Actual:** panel stays dark. Brightness keys do not recover it. ## Observed: max_brightness differs between the two kernels Measured on this machine: | Configuration | `max_brightness` | Resume | |---|---|---| | 7.0.0-27, stock cmdline | **1515** | works | | 7.0.0-28, stock cmdline | **1023** | backlight stays off | | 7.0.0-28 + `i915.enable_dpcd_backlight=0` | **1515** | works | `enable_dpcd_backlight` reads `-1` (the default) on 7.0.0-28 when the failure occurs. *Interpretation, not measured:* 1023 is 10-bit full scale and 1515 is not a power-of-two boundary, which suggests the two kernels are driving the backlight through different control paths, and that the parameter selects between them. I have not confirmed from the driver which path is active in either case. That inference rests only on the `max_brightness` values and on the parameter changing the outcome. ## The failure is silent No errors or warnings related to the failure appear in the journal for the affected boots. `PM: suspend exit` completes, and i915 and mei_hdcp rebind. Backlight sysfs during the failure, with the panel dark: ``` /sys/class/backlight/intel_backlight/brightness 1023 /sys/class/backlight/intel_backlight/actual_brightness 1023 /sys/class/backlight/intel_backlight/max_brightness 1023 /sys/class/backlight/intel_backlight/bl_power 0 ``` `/var/lib/systemd/backlight/pci-0000:00:02.0:backlight:intel_backlight` contained `1023`, so the saved brightness value was not lost or zeroed. ## Regression range `linux-image-generic-hwe-26.04` was upgraded 7.0.0-27.27 -> 7.0.0-28.28 on **2026-07-19 18:38 (UTC+1)**, unattended via packagekit (`Commandline: packagekit role='update-packages'` in `/var/log/apt/history.log`). 7.0.0-28 first booted 2026-07-22 15:14. The three-day gap is why the update was not initially suspected. From `journalctl --list-boots` and suspend entries per boot: | Kernel | Boots | Suspend entries | Outcome | |---|---|---|---| | 7.0.0-27 | 2 | 63 | boots lasted days, spanning many resumes | | 7.0.0-28 | 3 | 4 | each boot ended 39s to 2min after its first resume | The short sessions on 7.0.0-28 end in clean systemd shutdowns, from the machine being rebooted blind at a dark screen. ## Workaround Add to `GRUB_CMDLINE_LINUX_DEFAULT` in `/etc/default/grub`, then `update-grub`: ``` i915.enable_dpcd_backlight=0 ``` `max_brightness` returns to 1515 and resume works. Confirmed working after this change. I have not yet run an extended soak of many suspend cycles. ## Checked and excluded - **Not `systemd-backlight` save/restore:** the saved value was intact and sysfs read back at maximum while the panel was dark. - **Not local misconfiguration:** no i915 module parameters set, no backlight-related files in `/etc/modprobe.d`, and the cmdline was stock `quiet splash` before the workaround was added. - **A userspace resume hook did not help:** an `/etc/systemd/system-sleep/` hook that cycled `bl_power` 4 to 0 and rewrote `brightness` on resume did not light the panel. Caveat: I did not independently confirm the hook executed on resume, so this is weaker evidence than the rest of this report. ## Not verified Listing these so they are not read into the report: - Which backlight control path the driver actually selects in either kernel. - Whether the machine is network-reachable while the screen is dark. - The specific upstream commit responsible. I did not bisect, and the shipped `linux-image-7.0.0-28-generic` changelog contains only packaging entries, so I could not identify a candidate change from the package itself. ## Possibly related prior art There is a history of this general symptom with DPCD/AUX backlight on i915. A commit adding a heuristic to choose automatically between AUX and PWM brightness control was reverted upstream after reportedly causing "enabling DPCD backlight leads to a black screen after suspend/resume": https://www.mail-archive.com/intel- [email protected]/msg126115.html That report concerns different and much older hardware (a ThinkPad X1 Carbon 4th Gen and BXT-P), and predates this kernel by many releases. I offer it only as a pointer to a known failure mode in this area, not as a claim about the cause here. ProblemType: Bug DistroRelease: Ubuntu 26.04 Package: linux-image-7.0.0-28-generic 7.0.0-28.28 ProcVersionSignature: Ubuntu 7.0.0-28.28-generic 7.0.12 Uname: Linux 7.0.0-28-generic x86_64 ApportVersion: 2.34.0-0ubuntu2 Architecture: amd64 AudioDevicesInUse: USER PID ACCESS COMMAND /dev/snd/controlC0: sam 3370 F.... pipewire sam 3389 F.... wireplumber /dev/snd/pcmC0D0p: sam 3370 F...m pipewire /dev/snd/seq: sam 3370 F.... pipewire CasperMD5CheckResult: pass CurrentDesktop: KDE Date: Wed Jul 22 22:51:57 2026 InstallationDate: Installed on 2026-02-13 (159 days ago) InstallationMedia: Ubuntu 24.04.1 LTS "Noble Numbat" - Release amd64 (20240827.1) MachineType: LENOVO 20L6S0MD00 ProcFB: 0 i915drmfb ProcKernelCmdLine: BOOT_IMAGE=/vmlinuz-7.0.0-28-generic root=/dev/mapper/ubuntu--vg-ubuntu--lv ro quiet splash crashkernel=2G-4G:320M,4G-32G:512M,32G-64G:1024M,64G-128G:2048M,128G-:4096M i915.enable_dpcd_backlight=0 PulseList: Error: command ['pacmd', 'list'] failed with exit code 1: No PulseAudio daemon running, or not running as session daemon. SourcePackage: linux UpgradeStatus: Upgraded to resolute on 2026-07-09 (13 days ago) dmi.bios.date: 09/06/2025 dmi.bios.release: 1.56 dmi.bios.vendor: LENOVO dmi.bios.version: N24ET81W (1.56 ) dmi.board.asset.tag: Not Available dmi.board.name: 20L6S0MD00 dmi.board.vendor: LENOVO dmi.board.version: SDK0J40700 WIN dmi.chassis.asset.tag: No Asset Information dmi.chassis.type: 10 dmi.chassis.vendor: LENOVO dmi.chassis.version: None dmi.ec.firmware.release: 1.22 dmi.modalias: dmi:bvnLENOVO:bvrN24ET81W(1.56):bd09/06/2025:br1.56:efr1.22:svnLENOVO:pn20L6S0MD00:pvrThinkPadT480:rvnLENOVO:rn20L6S0MD00:rvrSDK0J40700WIN:cvnLENOVO:ct10:cvrNone:skuLENOVO_MT_20L6_BU_Think_FM_ThinkPadT480:pfaThinkPadT480: dmi.product.family: ThinkPad T480 dmi.product.name: 20L6S0MD00 dmi.product.sku: LENOVO_MT_20L6_BU_Think_FM_ThinkPad T480 dmi.product.version: ThinkPad T480 dmi.sys.vendor: LENOVO ** Affects: linux (Ubuntu) Importance: Undecided Status: New ** Tags: amd64 apport-bug resolute wayland-session -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2161575 Title: [regression] 7.0.0-28: eDP backlight stays off after resume on ThinkPad T480 (i915 selects DPCD/AUX backlight instead of native PWM) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2161575/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
