Public bug reported: Summary ------- The Lenovo ThinkPad P1 Gen 8 with NVIDIA RTX PRO 2000 (GB206GLM, 10de:2d39) is Ubuntu-certified (certificate 202510-37972, certified for 24.04 LTS + OEM kernel 6.14.0-1015-oem). On this same hardware running Ubuntu 26.04 / kernel 7.0.0-22-generic, the discrete GPU's GSP firmware crashes reproducibly 4-18 minutes after boot, leading to a continuous high-power state and hard power-off on battery.
Scope disclosure ---------------- I am running 26.04 + kernel 7.0, i.e. OUTSIDE the certified 24.04/6.14 combination. Reporting anyway because (a) it is the same certified hardware + GPU, (b) the root cause is an NVIDIA GSP-firmware regression that spans driver branches and is tracked upstream, and (c) Canonical/Lenovo OEM should be aware for the 26.04 cycle and for any BIOS/VBIOS mitigation. Hardware -------- ThinkPad P1 Gen 8, machine type 21Q9, BIOS N4EET22W (1.08). GPU: NVIDIA RTX PRO 2000 Blackwell Laptop (GB206GLM, 10de:2d39). iGPU Intel Arrow Lake-P drives the display. Symptom / kernel signature -------------------------- Xid 62 or Xid 120 "GSP task exception" (RISC-V cause varies between boots) -> Xid 154 "GPU recovery action ... GPU Reset Required" -> endless "RmCheckForGcxSupportOnCurrentState: Failed to get GCx pre-requisite, status=0x62" (every 5 s, up to 16879x/boot). The GPU never returns to a low-power GCx state, draws power continuously, and on battery the laptop hard-powers-off. nvidia-smi then reports "GPU requires reset" until reboot. Both NVIDIA driver branches affected (6 crash boots): nvidia- driver-595-open 595.71.05 AND nvidia-driver-580-open 580.159.03. Clean only when the nvidia module is not loaded (prime-select intel). Example (580.159.03): NVRM: Xid (PCI:0000:01:00): 120, GSP kernel exception: load access page fault (cause:0xd) @ pc:0xffffffff93008780, partition:4#0 NVRM: Xid (PCI:0000:01:00): 154, GPU recovery action changed from 0x0 (None) to 0x1 (GPU Reset Required) NVRM: RmCheckForGcxSupportOnCurrentState: NVRM, Failed to get GCx pre-requisite, status=0x62 Upstream / NVIDIA ----------------- Tracked in NVIDIA open-gpu-kernel-modules issue #1045 (NVIDIA internal bug 5953411). My data point there: https://github.com/NVIDIA/open-gpu-kernel-modules/issues/1045#issuecomment-4695950002 Workaround in place ------------------- prime-select intel disables the dGPU (module not loaded -> no GSP -> no crash); display runs on the Intel iGPU. Asks to Canonical OEM / Lenovo ------------------------------ 1. Confirm whether this reproduces on the certified 24.04 + 6.14-oem combo. 2. Is a BIOS/VBIOS/GSP-firmware update planned for the 21Q9 that addresses Blackwell GSP stability under Linux? 3. Track for the 26.04 certification cycle. ProblemType: Bug DistroRelease: Ubuntu 26.04 Package: linux 7.0.0-22-generic A sanitized nvidia-bug-report.log.gz can be attached on request. ** Affects: linux (Ubuntu) Importance: Undecided Status: New -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2156637 Title: [ThinkPad P1 Gen 8 / RTX PRO 2000 Blackwell] Reproducible NVIDIA GSP crash (Xid 120/154 + GCx loop), hard power-off on battery — certified hw 202510-37972 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2156637/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
