environment variables break suspend/resume mechanism (Ubuntu 25.04, nvidia driver 570)

[Jurgen-Schellaert]

After setting up the driver as documented on your github page, I found that my system (Ubuntu 25.04, nvidia 970) systematically failed to suspend. While the screen and input devices turned off, the rest of the system kept running.

It turns out that removing the environment variables is enough to fix this. For what it is worth, I mean these (which I set in /etc/environment.d):
LIBVA_DRIVER_NAME=nvidia
NVD_BACKEND=direct
MOZ_DISABLE_RDD_SANDBOX=1
NVD_LOG=/var/nvidia/vaapi.log

When the system fails to suspend, it logs this:
Jun 04 18:26:38 Dapple kernel: BUG: kernel NULL pointer dereference, address: 0000000000000000
Jun 04 18:26:38 Dapple kernel: fbcon: Taking over console
Jun 04 18:26:38 Dapple kernel: #PF: supervisor write access in kernel mode
Jun 04 18:26:38 Dapple kernel: #PF: error_code(0x0002) - not-present page
Jun 04 18:26:38 Dapple kernel: PGD 0 P4D 0
Jun 04 18:26:38 Dapple kernel: Oops: Oops: 0002 [#1] PREEMPT SMP NOPTI
Jun 04 18:26:38 Dapple kernel: CPU: 7 UID: 0 PID: 7877 Comm: nvidia-sleep.sh Tainted: P OE 6.14.0-15-generic #15-Ubuntu
Jun 04 18:26:38 Dapple kernel: Tainted: [P]=PROPRIETARY_MODULE, [O]=OOT_MODULE, [E]=UNSIGNED_MODULE
Jun 04 18:26:38 Dapple kernel: Hardware name: System manufacturer System Product Name/PRIME X470-PRO, BIOS 6232 09/29/2024
Jun 04 18:26:38 Dapple kernel: RIP: 0010:_raw_q_flush+0x92/0xf0 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: Code: 4c 89 64 24 18 e8 0e 34 e3 c4 48 89 c6 48 8b 04 24 48 39 e0 75 56 48 8b 53 08 48 89 1c 24 4c 89 f7 48 89 43 08 48 89 54 24 08 <48> 89 02 e8 86 34 e3 c4 48 8d 7b 1>
Jun 04 18:26:38 Dapple kernel: RSP: 0018:ffffa6b74e19b9e0 EFLAGS: 00010046
Jun 04 18:26:38 Dapple kernel: RAX: ffffa6b74e19b9e0 RBX: ffffa6b742f452f0 RCX: 0000000000000000
Jun 04 18:26:38 Dapple kernel: RDX: 0000000000000000 RSI: 0000000000000282 RDI: ffffa6b742f45300
Jun 04 18:26:38 Dapple kernel: RBP: ffffa6b74e19ba58 R08: 0000000000000000 R09: 0000000000000000
Jun 04 18:26:38 Dapple kernel: R10: 0000000000000000 R11: 0000000000000000 R12: ffffa6b74e19ba08
Jun 04 18:26:38 Dapple kernel: R13: ffffa6b742f45008 R14: ffffa6b742f45300 R15: ffff8fe00da6b000
Jun 04 18:26:38 Dapple kernel: FS: 00007b8f18cc8740(0000) GS:ffff8fe64ef80000(0000) knlGS:0000000000000000
Jun 04 18:26:38 Dapple kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
Jun 04 18:26:38 Dapple kernel: CR2: 0000000000000000 CR3: 00000001fe628000 CR4: 0000000000350ef0
Jun 04 18:26:38 Dapple kernel: Call Trace:
Jun 04 18:26:38 Dapple kernel:
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? show_trace_log_lvl+0x273/0x310
Jun 04 18:26:38 Dapple kernel: ? show_trace_log_lvl+0x273/0x310
Jun 04 18:26:38 Dapple kernel: ? nv_kthread_q_flush+0x19/0x80 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: ? show_regs.part.0+0x22/0x30
Jun 04 18:26:38 Dapple kernel: ? __die_body.cold+0x8/0x10
Jun 04 18:26:38 Dapple kernel: ? __die+0x2a/0x40
Jun 04 18:26:38 Dapple kernel: ? page_fault_oops+0x16e/0x180
Jun 04 18:26:38 Dapple kernel: ? do_user_addr_fault+0x4c9/0x7e0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? __schedule+0x2ca/0x660
Jun 04 18:26:38 Dapple kernel: ? exc_page_fault+0x85/0x1e0
Jun 04 18:26:38 Dapple kernel: ? asm_exc_page_fault+0x27/0x30
Jun 04 18:26:38 Dapple kernel: ? _raw_q_flush+0x92/0xf0 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: ? __pfx__q_flush_function+0x10/0x10 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: ? thread_context_current+0x8e/0x130 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: nv_kthread_q_flush+0x19/0x80 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: uvm_suspend.isra.0+0x9e/0x1f0 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: uvm_suspend_entry+0x83/0xa0 [nvidia_uvm]
Jun 04 18:26:38 Dapple kernel: nv_uvm_suspend+0x31/0x50 [nvidia]
Jun 04 18:26:38 Dapple kernel: nv_set_system_power_state+0x3e2/0x490 [nvidia]
Jun 04 18:26:38 Dapple kernel: nv_procfs_write_suspend+0x102/0x1b0 [nvidia]
Jun 04 18:26:38 Dapple kernel: proc_reg_write+0x69/0xa0
Jun 04 18:26:38 Dapple kernel: vfs_write+0x10a/0x460
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? try_charge_memcg+0x562/0x5c0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? get_close_on_exec+0x34/0x50
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? do_fcntl+0x2ef/0x640
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? set_close_on_exec+0x4c/0x80
Jun 04 18:26:38 Dapple kernel: ksys_write+0x70/0xf0
Jun 04 18:26:38 Dapple kernel: __x64_sys_write+0x19/0x30
Jun 04 18:26:38 Dapple kernel: x64_sys_call+0x2a3/0x2310
Jun 04 18:26:38 Dapple kernel: do_syscall_64+0x7e/0x170
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? filp_flush+0x5a/0xb0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? arch_exit_to_user_mode_prepare.isra.0+0x22/0xd0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? syscall_exit_to_user_mode+0x38/0x1d0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? do_syscall_64+0x8a/0x170
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? ksys_dup3+0x9d/0x120
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? arch_exit_to_user_mode_prepare.isra.0+0x22/0xd0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? syscall_exit_to_user_mode+0x38/0x1d0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? do_syscall_64+0x8a/0x170
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? arch_exit_to_user_mode_prepare.isra.0+0x22/0xd0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? irqentry_exit_to_user_mode+0x2d/0x1d0
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? irqentry_exit+0x43/0x50
Jun 04 18:26:38 Dapple kernel: ? srso_return_thunk+0x5/0x5f
Jun 04 18:26:38 Dapple kernel: ? exc_page_fault+0x96/0x1e0
Jun 04 18:26:38 Dapple kernel: entry_SYSCALL_64_after_hwframe+0x76/0x7e
Jun 04 18:26:38 Dapple kernel: RIP: 0033:0x7b8f18a9f2a6
Jun 04 18:26:38 Dapple kernel: Code: 00 00 48 8b 15 53 1b 17 00 64 89 02 48 c7 c2 ff ff ff ff 48 8b 5d f8 c9 48 89 d0 c3 0f 1f 84 00 00 00 00 00 48 8b 45 10 0f 05 <48> 63 d0 3d 00 f0 ff ff 77 10 48 8>
Jun 04 18:26:38 Dapple kernel: RSP: 002b:00007ffc502217d0 EFLAGS: 00000202 ORIG_RAX: 0000000000000001
Jun 04 18:26:38 Dapple kernel: RAX: ffffffffffffffda RBX: 00007b8f18cc8740 RCX: 00007b8f18a9f2a6
Jun 04 18:26:38 Dapple kernel: RDX: 0000000000000008 RSI: 000057adeff140d0 RDI: 0000000000000001
Jun 04 18:26:38 Dapple kernel: note: nvidia-sleep.sh[7877] exited with irqs disabled
Jun 04 18:26:38 Dapple kernel: note: nvidia-sleep.sh[7877] exited with preempt_count 1
Jun 04 18:26:38 Dapple kernel: Console: switching to colour frame buffer device 240x67
Jun 04 18:26:38 Dapple systemd[1]: nvidia-suspend.service: Main process exited, code=killed, status=9/KILL
Jun 04 18:26:38 Dapple systemd[1]: nvidia-suspend.service: Failed with result 'signal'.
Jun 04 18:26:38 Dapple systemd[1]: Failed to start nvidia-suspend.service - NVIDIA system suspend actions.

[elFarto]

There's not much I can do about a crash in the nvidia_uvm module. Might be better to raise this on the NVIDIA forums.

I don't think it's the environment variables per-se that's breaking your system, but the fact that we poke around in the driver. One thing I would be interested to know, does this crash occur if you start the system with those variable set, and suspend before this driver is ever run? In theory we wouldn't have touched the driver in anyway so it should suspend just fine.

[EiPiFun]

There's not much I can do about a crash in the nvidia_uvm module. Might be better to raise this on the NVIDIA forums.

I don't think it's the environment variables per-se that's breaking your system, but the fact that we poke around in the driver. One thing I would be interested to know, does this crash occur if you start the system with those variable set, and suspend before this driver is ever run? In theory we wouldn't have touched the driver in anyway so it should suspend just fine.

I have met similar problems. In my case, only when NV GPU is used for hardware decoding, suspend will break.
Setting LIBVA_DRIVER_NAME=nv (or any other nonsense) will recover the suspend/resume.
Using mpv (hwdec=vulkan) will not break suspend/resume.
I guess it is something about suspend policy when CUDA computing task is performing.