[Nvidia-dkms-595.84] Missing ENDBR in init_module causes kernel BUG on IBT/CET-enabled kernel (Xanmod Clang)

[MinecraftSky]

NVIDIA Open GPU Kernel Modules Version

595.84 (Installed via Ubuntu DKMS package)

Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.

• I confirm that this does not happen with the proprietary driver package.

Operating System and Version

Description: Ubuntu 26.04 LTS

Kernel Release

Linux Nitro-NL16-71G 7.1.4-x64v3-xanmod1 #0~20260718.g967cc06 SMP PREEMPT_DYNAMIC Sat Jul 18 17:42:15 UTC x86_64 GNU/Linux

Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.

• I am running on a stable kernel release.

Hardware: GPU

GPU 0: NVIDIA GeForce RTX 4050 Laptop GPU (UUID: GPU-4096e3ec-fcb2-bd62-c2c3-3f7e49e2ad14)

Describe the bug

When booting a Linux kernel with Indirect Branch Tracking (IBT) strictly enabled (CET detected: Indirect Branch Tracking enabled), loading the nvidia kernel module triggers a hardware control protection fault (do_kernel_cp_fault) and hangs the system.

This occurs because the entry point (init_module) in the NVIDIA kernel module lacks the ENDBR instruction required by Intel CET/IBT.

Environment

• GPU: NVIDIA GeForce RTX 4050 Laptop GPU (Acer Nitro NL16-71G)
• CPU: Intel Core i5-13420H (CET/IBT support)
• Driver Version: 595.84
• Kernel Version: 7.1.4-x64v3-xanmod1 (Compiled with Clang/LLD, CONFIG_X86_KERNEL_IBT=y)
• OS: Ubuntu 26.04 LTS

---

Key Log Snippet

kernel: CET detected: Indirect Branch Tracking enabled
kernel: Missing ENDBR: init_module+0x0/0x340 [nvidia]
kernel: kernel BUG at do_kernel_cp_fault+0xd7/0xe0

---

To Reproduce

Preconditions:

• A CPU supporting Intel CET / Indirect Branch Tracking (IBT) (e.g., Intel 11th Gen or newer).
• A Linux kernel compiled with CONFIG_X86_KERNEL_IBT=y (e.g., Xanmod Linux kernel).
• NVIDIA Open GPU Kernel Modules installed via DKMS.

Steps to Reproduce:

1. Configure GRUB boot parameters without ibt=off.
2. Boot into the IBT-enabled kernel.
3. Allow the system to boot and attempt loading the nvidia kernel module.
4. Observe system freeze / soft lockup on boot, accompanied by Missing ENDBR: init_module in kernel logs.

---

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz
error-kernel-log.zip

More Info

Expected Behavior:
The NVIDIA kernel module should load successfully during boot when Intel CET/IBT is enabled (CONFIG_X86_KERNEL_IBT=y), allowing the system to boot into the desktop environment with full GPU hardware acceleration.

Actual Behavior:
The system triggers a kernel BUG (do_kernel_cp_fault) and completely freezes (soft lockup) because the init_module entry point in the NVIDIA driver lacks the required ENDBR instruction for IBT validation.

Additional Context:

• Passing ibt=off as a kernel boot parameter bypasses the check and works around the crash completely.
• Attached nvidia-bug-report.log.gz for full system diagnosis.

[MinecraftSky]

Related to prior issue — this is a recurring, not new, problem:

This appears to be the same root cause as #256 (Linux 5.18 NVIDIA module
won't load: Missing ENDBR: _portMemAllocatorAllocNonPagedWrapper+0x0/0x10),
first reported in 2022 when IBT support landed in kernel 5.18.

Four years later, with driver 595.84 and kernel 7.1.4-x64v3-xanmod1, the same
Missing ENDBR symptom still occurs — this time in init_module rather than
_portMemAllocatorAllocNonPagedWrapper, suggesting the underlying build/CFLAGS
issue (ENDBR not being emitted for CET/IBT-enabled kernels) has never been
fully resolved, only worked around via ibt=off.

---

Clarification on proprietary driver confirmation:

I marked "I confirm that this does not happen with the proprietary driver package"
based on the assumption that this is an open-kernel-module-specific build/linking
issue (missing ENDBR from CFLAGS mismatch), rather than from direct testing on
this machine.

I have not personally reproduced this bug with the proprietary (closed-source)
driver on this system. However, based on prior reports (e.g. #256, from kernel
5.18 in 2022), this same Missing ENDBR symptom was also observed with the
proprietary driver at the time — so I cannot rule out that this affects both
flavors.

If maintainers need it, I'm happy to test with the proprietary driver package
to confirm either way.