Niri cannot be frozen when session locked

[RMTT]

Recently i'm using noctalia-shell(based on quickshell) and swayidle, and following is my swayidle's configuration:

  services.swayidle = {
    enable = true;
    timeouts = [
      {
        timeout = 600;
        command = "${pkgs.noctalia}/bin/noctalia-shell ipc call lockScreen toggle";
      }
      {
        timeout = 1800;
        command = "${pkgs.niri}/bin/niri msg action power-off-monitors";
      }
    ];
  };

It's configured by home-manager and it's very simple.
After the desktop remains idle for 10 minutes, invoking noctalia-shell ipc call lockScreen toggle to open a lock screen ui;
After the desktop remains idle for 30 minutes, invoking niri msg action power-off-monitors to power-off monitors;

After niri msg action power-off-monitors, systemd-logind will start suspending and an error will occur:

Sep 24 15:01:19 mtspc systemd-sleep[71623]: Failed to freeze unit 'user.slice': Connection timed out
Sep 24 15:01:19 mtspc systemd-sleep[71623]: Performing sleep operation 'suspend'...
Sep 24 15:01:19 mtspc kernel: PM: suspend entry (s2idle)
Sep 24 15:01:19 mtspc kernel: Filesystems sync: 0.008 seconds
Sep 24 15:01:39 mtspc kernel: Freezing user space processes
Sep 24 15:01:39 mtspc kernel: Freezing user space processes failed after 20.007 seconds (1 tasks refusing to freeze, wq_busy=0):
Sep 24 15:01:39 mtspc kernel: task:niri            state:D stack:0     pid:2348  tgid:2348  ppid:2293   task_flags:0x400000 flags:0x00004006
Sep 24 15:01:39 mtspc kernel: Call Trace:
Sep 24 15:01:39 mtspc kernel:  <TASK>
Sep 24 15:01:39 mtspc kernel:  __schedule+0x465/0x12f0
Sep 24 15:01:39 mtspc kernel:  schedule+0x27/0xd0
Sep 24 15:01:39 mtspc kernel:  schedule_preempt_disabled+0x15/0x30
Sep 24 15:01:39 mtspc kernel:  __mutex_lock.constprop.0+0x3f5/0x7d0
Sep 24 15:01:39 mtspc kernel:  drm_mode_getconnector+0x11e/0x480
Sep 24 15:01:39 mtspc kernel:  ? __pfx_drm_mode_getconnector+0x10/0x10
Sep 24 15:01:39 mtspc kernel:  drm_ioctl_kernel+0xab/0x100
Sep 24 15:01:39 mtspc kernel:  drm_ioctl+0x273/0x560
Sep 24 15:01:39 mtspc kernel:  ? __pfx_drm_mode_getconnector+0x10/0x10
Sep 24 15:01:39 mtspc kernel:  __x64_sys_ioctl+0x93/0xe0
Sep 24 15:01:39 mtspc kernel:  do_syscall_64+0xb7/0x370
Sep 24 15:01:39 mtspc kernel:  entry_SYSCALL_64_after_hwframe+0x77/0x7f
Sep 24 15:01:39 mtspc kernel: RIP: 0033:0x55736dc17d0e
Sep 24 15:01:39 mtspc kernel: RSP: 002b:00007ffd8adc03d0 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
Sep 24 15:01:39 mtspc kernel: RAX: ffffffffffffffda RBX: 000000000000002d RCX: 000055736dc17d0e
Sep 24 15:01:39 mtspc kernel: RDX: 00007ffd8adc0440 RSI: 00000000c05064a7 RDI: 000000000000002d
Sep 24 15:01:39 mtspc kernel: RBP: 0000000000000000 R08: 00007ffd8adc05a0 R09: 0000000000000080
Sep 24 15:01:39 mtspc kernel: R10: 00007ffd8adc0570 R11: 0000000000000246 R12: 00007ffd8adc0588
Sep 24 15:01:39 mtspc kernel: R13: 00005573a0008a10 R14: 0000000000000000 R15: 0000000000000000
Sep 24 15:01:39 mtspc kernel:  </TASK>
Sep 24 15:01:39 mtspc kernel: OOM killer enabled.
Sep 24 15:01:39 mtspc kernel: Restarting tasks: Starting
Sep 24 15:01:39 mtspc kernel: Restarting tasks: Done
Sep 24 15:01:39 mtspc kernel: random: crng reseeded on system resumption

I noticed that quickshell implemented ext-session-lock, when lock screen toggled, niri will update LockedHint of Logind.

Any thoughts for the root cause?

[YaLTeR]

What does this backtrace mean? Why do you think niri is at fault (and if so, then for what)?

[RMTT]

Why do you think niri is at fault (and if so, then for what)?

Hmm, I'm not sure whether it's niri's bug. Maybe someone encountered the same issue and has a solution, so I posted those information here.

What does this backtrace mean?

Seems it's a backtrace after niri invokes drm syscall. I'm not familiar with kernel code, but i guess that the code path is niri poweroff monitors -> ... -> invoke drm syscall.

I think it's probably a bug of nvidia driver. Sorry to bother you, i will close this issue.