1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
|
{
config,
lib,
pkgs,
...
}:
let
nvidiaEnabled = (lib.elem "nvidia" config.services.xserver.videoDrivers);
nvidia_x11 = if nvidiaEnabled || cfg.datacenter.enable then cfg.package else null;
cfg = config.hardware.nvidia;
pCfg = cfg.prime;
syncCfg = pCfg.sync;
offloadCfg = pCfg.offload;
reverseSyncCfg = pCfg.reverseSync;
primeEnabled = syncCfg.enable || reverseSyncCfg.enable || offloadCfg.enable;
busIDType = lib.types.strMatching "([[:print:]]+[\:\@][0-9]{1,3}\:[0-9]{1,2}\:[0-9])?";
ibtSupport = cfg.open || (nvidia_x11.ibtSupport or false);
settingsFormat = pkgs.formats.keyValue { };
in
{
options = {
hardware.nvidia = {
datacenter.enable = lib.mkEnableOption ''
Data Center drivers for NVIDIA cards on a NVLink topology
'';
datacenter.settings = lib.mkOption {
type = settingsFormat.type;
default = {
LOG_LEVEL = 4;
LOG_FILE_NAME = "/var/log/fabricmanager.log";
LOG_APPEND_TO_LOG = 1;
LOG_FILE_MAX_SIZE = 1024;
LOG_USE_SYSLOG = 0;
DAEMONIZE = 1;
BIND_INTERFACE_IP = "127.0.0.1";
STARTING_TCP_PORT = 16000;
FABRIC_MODE = 0;
FABRIC_MODE_RESTART = 0;
STATE_FILE_NAME = "/var/tmp/fabricmanager.state";
FM_CMD_BIND_INTERFACE = "127.0.0.1";
FM_CMD_PORT_NUMBER = 6666;
FM_STAY_RESIDENT_ON_FAILURES = 0;
ACCESS_LINK_FAILURE_MODE = 0;
TRUNK_LINK_FAILURE_MODE = 0;
NVSWITCH_FAILURE_MODE = 0;
ABORT_CUDA_JOBS_ON_FM_EXIT = 1;
};
defaultText = lib.literalExpression ''
{
LOG_LEVEL=4;
LOG_FILE_NAME="/var/log/fabricmanager.log";
LOG_APPEND_TO_LOG=1;
LOG_FILE_MAX_SIZE=1024;
LOG_USE_SYSLOG=0;
DAEMONIZE=1;
BIND_INTERFACE_IP="127.0.0.1";
STARTING_TCP_PORT=16000;
FABRIC_MODE=0;
FABRIC_MODE_RESTART=0;
STATE_FILE_NAME="/var/tmp/fabricmanager.state";
FM_CMD_BIND_INTERFACE="127.0.0.1";
FM_CMD_PORT_NUMBER=6666;
FM_STAY_RESIDENT_ON_FAILURES=0;
ACCESS_LINK_FAILURE_MODE=0;
TRUNK_LINK_FAILURE_MODE=0;
NVSWITCH_FAILURE_MODE=0;
ABORT_CUDA_JOBS_ON_FM_EXIT=1;
}
'';
description = ''
Additional configuration options for fabricmanager.
'';
};
powerManagement.enable = lib.mkEnableOption ''
experimental power management through systemd. For more information, see
the NVIDIA docs, on Chapter 21. Configuring Power Management Support
'';
powerManagement.finegrained = lib.mkEnableOption ''
experimental power management of PRIME offload. For more information, see
the NVIDIA docs, on Chapter 22. PCI-Express Runtime D3 (RTD3) Power Management
'';
dynamicBoost.enable = lib.mkEnableOption ''
dynamic Boost balances power between the CPU and the GPU for improved
performance on supported laptops using the nvidia-powerd daemon. For more
information, see the NVIDIA docs, on Chapter 23. Dynamic Boost on Linux
'';
modesetting.enable = lib.mkEnableOption ''
kernel modesetting when using the NVIDIA proprietary driver.
Enabling this fixes screen tearing when using Optimus via PRIME (see
{option}`hardware.nvidia.prime.sync.enable`. This is not enabled
by default because it is not officially supported by NVIDIA and would not
work with SLI.
Enabling this and using version 545 or newer of the proprietary NVIDIA
driver causes it to provide its own framebuffer device, which can cause
Wayland compositors to work when they otherwise wouldn't.
'' // {
default = lib.versionAtLeast nvidia_x11.version "535";
defaultText = lib.literalExpression "lib.versionAtLeast nvidia_x11.version \"535\"";
};
prime.nvidiaBusId = lib.mkOption {
type = busIDType;
default = "";
example = "PCI:1:0:0";
description = ''
Bus ID of the NVIDIA GPU. You can find it using lspci; for example if lspci
shows the NVIDIA GPU at "01:00.0", set this option to "PCI:1:0:0".
'';
};
prime.intelBusId = lib.mkOption {
type = busIDType;
default = "";
example = "PCI:0:2:0";
description = ''
Bus ID of the Intel GPU. You can find it using lspci; for example if lspci
shows the Intel GPU at "00:02.0", set this option to "PCI:0:2:0".
'';
};
prime.amdgpuBusId = lib.mkOption {
type = busIDType;
default = "";
example = "PCI:4:0:0";
description = ''
Bus ID of the AMD APU. You can find it using lspci; for example if lspci
shows the AMD APU at "04:00.0", set this option to "PCI:4:0:0".
'';
};
prime.sync.enable = lib.mkEnableOption ''
NVIDIA Optimus support using the NVIDIA proprietary driver via PRIME.
If enabled, the NVIDIA GPU will be always on and used for all rendering,
while enabling output to displays attached only to the integrated Intel/AMD
GPU without a multiplexer.
Note that this option only has any effect if the "nvidia" driver is specified
in {option}`services.xserver.videoDrivers`, and it should preferably
be the only driver there.
If this is enabled, then the bus IDs of the NVIDIA and Intel/AMD GPUs have to
be specified ({option}`hardware.nvidia.prime.nvidiaBusId` and
{option}`hardware.nvidia.prime.intelBusId` or
{option}`hardware.nvidia.prime.amdgpuBusId`).
If you enable this, you may want to also enable kernel modesetting for the
NVIDIA driver ({option}`hardware.nvidia.modesetting.enable`) in order
to prevent tearing.
Note that this configuration will only be successful when a display manager
for which the {option}`services.xserver.displayManager.setupCommands`
option is supported is used
'';
prime.allowExternalGpu = lib.mkEnableOption ''
configuring X to allow external NVIDIA GPUs when using Prime [Reverse] sync optimus
'';
prime.offload.enable = lib.mkEnableOption ''
render offload support using the NVIDIA proprietary driver via PRIME.
If this is enabled, then the bus IDs of the NVIDIA and Intel/AMD GPUs have to
be specified ({option}`hardware.nvidia.prime.nvidiaBusId` and
{option}`hardware.nvidia.prime.intelBusId` or
{option}`hardware.nvidia.prime.amdgpuBusId`)
'';
prime.offload.enableOffloadCmd = lib.mkEnableOption ''
adding a `nvidia-offload` convenience script to {option}`environment.systemPackages`
for offloading programs to an nvidia device. To work, should have also enabled
{option}`hardware.nvidia.prime.offload.enable` or {option}`hardware.nvidia.prime.reverseSync.enable`.
Example usage `nvidia-offload sauerbraten_client`
'';
prime.reverseSync.enable = lib.mkEnableOption ''
NVIDIA Optimus support using the NVIDIA proprietary driver via reverse
PRIME. If enabled, the Intel/AMD GPU will be used for all rendering, while
enabling output to displays attached only to the NVIDIA GPU without a
multiplexer.
Warning: This feature is relatively new, depending on your system this might
work poorly. AMD support, especially so.
See: https://forums.developer.nvidia.com/t/the-all-new-outputsink-feature-aka-reverse-prime/129828
Note that this option only has any effect if the "nvidia" driver is specified
in {option}`services.xserver.videoDrivers`, and it should preferably
be the only driver there.
If this is enabled, then the bus IDs of the NVIDIA and Intel/AMD GPUs have to
be specified ({option}`hardware.nvidia.prime.nvidiaBusId` and
{option}`hardware.nvidia.prime.intelBusId` or
{option}`hardware.nvidia.prime.amdgpuBusId`).
If you enable this, you may want to also enable kernel modesetting for the
NVIDIA driver ({option}`hardware.nvidia.modesetting.enable`) in order
to prevent tearing.
Note that this configuration will only be successful when a display manager
for which the {option}`services.xserver.displayManager.setupCommands`
option is supported is used
'';
prime.reverseSync.setupCommands.enable =
(lib.mkEnableOption ''
configure the display manager to be able to use the outputs
attached to the NVIDIA GPU.
Disable in order to configure the NVIDIA GPU outputs manually using xrandr.
Note that this configuration will only be successful when a display manager
for which the {option}`services.xserver.displayManager.setupCommands`
option is supported is used
'')
// {
default = true;
};
nvidiaSettings =
(lib.mkEnableOption ''
nvidia-settings, NVIDIA's GUI configuration tool
'')
// {
default = true;
};
nvidiaPersistenced = lib.mkEnableOption ''
nvidia-persistenced a update for NVIDIA GPU headless mode, i.e.
It ensures all GPUs stay awake even during headless mode
'';
forceFullCompositionPipeline = lib.mkEnableOption ''
forcefully the full composition pipeline.
This sometimes fixes screen tearing issues.
This has been reported to reduce the performance of some OpenGL applications and may produce issues in WebGL.
It also drastically increases the time the driver needs to clock down after load
'';
package = lib.mkOption {
default =
config.boot.kernelPackages.nvidiaPackages."${if cfg.datacenter.enable then "dc" else "stable"}";
defaultText = lib.literalExpression ''
config.boot.kernelPackages.nvidiaPackages."\$\{if cfg.datacenter.enable then "dc" else "stable"}"
'';
example = "config.boot.kernelPackages.nvidiaPackages.legacy_470";
description = ''
The NVIDIA driver package to use.
'';
};
open = lib.mkEnableOption ''
the open source NVIDIA kernel module
'';
};
};
config =
let
igpuDriver = if pCfg.intelBusId != "" then "modesetting" else "amdgpu";
igpuBusId = if pCfg.intelBusId != "" then pCfg.intelBusId else pCfg.amdgpuBusId;
in
lib.mkIf (nvidia_x11 != null) (
lib.mkMerge [
# Common
({
assertions = [
{
assertion = !(nvidiaEnabled && cfg.datacenter.enable);
message = "You cannot configure both X11 and Data Center drivers at the same time.";
}
];
boot = {
blacklistedKernelModules = [
"nouveau"
"nvidiafb"
];
# Don't add `nvidia-uvm` to `kernelModules`, because we want
# `nvidia-uvm` be loaded only after `udev` rules for `nvidia` kernel
# module are applied.
#
# Instead, we use `softdep` to lazily load `nvidia-uvm` kernel module
# after `nvidia` kernel module is loaded and `udev` rules are applied.
extraModprobeConfig = ''
softdep nvidia post: nvidia-uvm
'';
};
systemd.tmpfiles.rules = lib.mkIf config.virtualisation.docker.enableNvidia [ "L+ /run/nvidia-docker/bin - - - - ${nvidia_x11.bin}/origBin" ];
services.udev.extraRules = ''
# Create /dev/nvidia-uvm when the nvidia-uvm module is loaded.
KERNEL=="nvidia", RUN+="${pkgs.runtimeShell} -c 'mknod -m 666 /dev/nvidiactl c 195 255'"
KERNEL=="nvidia", RUN+="${pkgs.runtimeShell} -c 'for i in $$(cat /proc/driver/nvidia/gpus/*/information | grep Minor | cut -d \ -f 4); do mknod -m 666 /dev/nvidia$${i} c 195 $${i}; done'"
KERNEL=="nvidia_modeset", RUN+="${pkgs.runtimeShell} -c 'mknod -m 666 /dev/nvidia-modeset c 195 254'"
KERNEL=="nvidia_uvm", RUN+="${pkgs.runtimeShell} -c 'mknod -m 666 /dev/nvidia-uvm c $$(grep nvidia-uvm /proc/devices | cut -d \ -f 1) 0'"
KERNEL=="nvidia_uvm", RUN+="${pkgs.runtimeShell} -c 'mknod -m 666 /dev/nvidia-uvm-tools c $$(grep nvidia-uvm /proc/devices | cut -d \ -f 1) 1'"
'';
hardware.graphics = {
extraPackages = [ nvidia_x11.out ];
extraPackages32 = [ nvidia_x11.lib32 ];
};
environment.systemPackages = [ nvidia_x11.bin ];
})
# X11
(lib.mkIf nvidiaEnabled {
assertions = [
{
assertion = primeEnabled -> pCfg.intelBusId == "" || pCfg.amdgpuBusId == "";
message = "You cannot configure both an Intel iGPU and an AMD APU. Pick the one corresponding to your processor.";
}
{
assertion = offloadCfg.enableOffloadCmd -> offloadCfg.enable || reverseSyncCfg.enable;
message = "Offload command requires offloading or reverse prime sync to be enabled.";
}
{
assertion =
primeEnabled -> pCfg.nvidiaBusId != "" && (pCfg.intelBusId != "" || pCfg.amdgpuBusId != "");
message = "When NVIDIA PRIME is enabled, the GPU bus IDs must be configured.";
}
{
assertion = offloadCfg.enable -> lib.versionAtLeast nvidia_x11.version "435.21";
message = "NVIDIA PRIME render offload is currently only supported on versions >= 435.21.";
}
{
assertion =
(reverseSyncCfg.enable && pCfg.amdgpuBusId != "") -> lib.versionAtLeast nvidia_x11.version "470.0";
message = "NVIDIA PRIME render offload for AMD APUs is currently only supported on versions >= 470 beta.";
}
{
assertion = !(syncCfg.enable && offloadCfg.enable);
message = "PRIME Sync and Offload cannot be both enabled";
}
{
assertion = !(syncCfg.enable && reverseSyncCfg.enable);
message = "PRIME Sync and PRIME Reverse Sync cannot be both enabled";
}
{
assertion = !(syncCfg.enable && cfg.powerManagement.finegrained);
message = "Sync precludes powering down the NVIDIA GPU.";
}
{
assertion = cfg.powerManagement.finegrained -> offloadCfg.enable;
message = "Fine-grained power management requires offload to be enabled.";
}
{
assertion = cfg.powerManagement.enable -> lib.versionAtLeast nvidia_x11.version "430.09";
message = "Required files for driver based power management only exist on versions >= 430.09.";
}
{
assertion = cfg.open -> (cfg.package ? open && cfg.package ? firmware);
message = "This version of NVIDIA driver does not provide a corresponding opensource kernel driver";
}
{
assertion = cfg.dynamicBoost.enable -> lib.versionAtLeast nvidia_x11.version "510.39.01";
message = "NVIDIA's Dynamic Boost feature only exists on versions >= 510.39.01";
}
];
# If Optimus/PRIME is enabled, we:
# - Specify the configured NVIDIA GPU bus ID in the Device section for the
# "nvidia" driver.
# - Add the AllowEmptyInitialConfiguration option to the Screen section for the
# "nvidia" driver, in order to allow the X server to start without any outputs.
# - Add a separate Device section for the Intel GPU, using the "modesetting"
# driver and with the configured BusID.
# - OR add a separate Device section for the AMD APU, using the "amdgpu"
# driver and with the configures BusID.
# - Reference that Device section from the ServerLayout section as an inactive
# device.
# - Configure the display manager to run specific `xrandr` commands which will
# configure/enable displays connected to the Intel iGPU / AMD APU.
# reverse sync implies offloading
hardware.nvidia.prime.offload.enable = lib.mkDefault reverseSyncCfg.enable;
services.xserver.drivers =
lib.optional primeEnabled {
name = igpuDriver;
display = offloadCfg.enable;
modules = lib.optional (igpuDriver == "amdgpu") pkgs.xorg.xf86videoamdgpu;
deviceSection =
''
BusID "${igpuBusId}"
''
+ lib.optionalString (syncCfg.enable && igpuDriver != "amdgpu") ''
Option "AccelMethod" "none"
'';
}
++ lib.singleton {
name = "nvidia";
modules = [ nvidia_x11.bin ];
display = !offloadCfg.enable;
deviceSection =
''
Option "SidebandSocketPath" "/run/nvidia-xdriver/"
''
+ lib.optionalString primeEnabled ''
BusID "${pCfg.nvidiaBusId}"
''
+ lib.optionalString pCfg.allowExternalGpu ''
Option "AllowExternalGpus"
'';
screenSection =
''
Option "RandRRotation" "on"
''
+ lib.optionalString syncCfg.enable ''
Option "AllowEmptyInitialConfiguration"
''
+ lib.optionalString cfg.forceFullCompositionPipeline ''
Option "metamodes" "nvidia-auto-select +0+0 {ForceFullCompositionPipeline=On}"
Option "AllowIndirectGLXProtocol" "off"
Option "TripleBuffer" "on"
'';
};
services.xserver.serverLayoutSection =
lib.optionalString syncCfg.enable ''
Inactive "Device-${igpuDriver}[0]"
''
+ lib.optionalString reverseSyncCfg.enable ''
Inactive "Device-nvidia[0]"
''
+ lib.optionalString offloadCfg.enable ''
Option "AllowNVIDIAGPUScreens"
'';
services.xserver.displayManager.setupCommands =
let
gpuProviderName =
if igpuDriver == "amdgpu" then
# find the name of the provider if amdgpu
"`${lib.getExe pkgs.xorg.xrandr} --listproviders | ${lib.getExe pkgs.gnugrep} -i AMD | ${lib.getExe pkgs.gnused} -n 's/^.*name://p'`"
else
igpuDriver;
providerCmdParams =
if syncCfg.enable then "\"${gpuProviderName}\" NVIDIA-0" else "NVIDIA-G0 \"${gpuProviderName}\"";
in
lib.optionalString
(syncCfg.enable || (reverseSyncCfg.enable && reverseSyncCfg.setupCommands.enable))
''
# Added by nvidia configuration module for Optimus/PRIME.
${lib.getExe pkgs.xorg.xrandr} --setprovideroutputsource ${providerCmdParams}
${lib.getExe pkgs.xorg.xrandr} --auto
'';
environment.etc = {
"nvidia/nvidia-application-profiles-rc" = lib.mkIf nvidia_x11.useProfiles {
source = "${nvidia_x11.bin}/share/nvidia/nvidia-application-profiles-rc";
};
# 'nvidia_x11' installs it's files to /run/opengl-driver/...
"egl/egl_external_platform.d".source = "/run/opengl-driver/share/egl/egl_external_platform.d/";
};
hardware.graphics = {
extraPackages = [ pkgs.nvidia-vaapi-driver ];
};
environment.systemPackages =
lib.optional cfg.nvidiaSettings nvidia_x11.settings
++ lib.optional cfg.nvidiaPersistenced nvidia_x11.persistenced
++ lib.optional offloadCfg.enableOffloadCmd (
pkgs.writeShellScriptBin "nvidia-offload" ''
export __NV_PRIME_RENDER_OFFLOAD=1
export __NV_PRIME_RENDER_OFFLOAD_PROVIDER=NVIDIA-G0
export __GLX_VENDOR_LIBRARY_NAME=nvidia
export __VK_LAYER_NV_optimus=NVIDIA_only
exec "$@"
''
);
systemd.packages = lib.optional cfg.powerManagement.enable nvidia_x11.out;
systemd.services =
let
nvidiaService = state: {
description = "NVIDIA system ${state} actions";
path = [ pkgs.kbd ];
serviceConfig = {
Type = "oneshot";
ExecStart = "${nvidia_x11.out}/bin/nvidia-sleep.sh '${state}'";
};
before = [ "systemd-${state}.service" ];
requiredBy = [ "systemd-${state}.service" ];
};
in
lib.mkMerge [
(lib.mkIf cfg.powerManagement.enable {
nvidia-suspend = nvidiaService "suspend";
nvidia-hibernate = nvidiaService "hibernate";
nvidia-resume = (nvidiaService "resume") // {
before = [ ];
after = [
"systemd-suspend.service"
"systemd-hibernate.service"
];
requiredBy = [
"systemd-suspend.service"
"systemd-hibernate.service"
];
};
})
(lib.mkIf cfg.nvidiaPersistenced {
"nvidia-persistenced" = {
description = "NVIDIA Persistence Daemon";
wantedBy = [ "multi-user.target" ];
serviceConfig = {
Type = "forking";
Restart = "always";
PIDFile = "/var/run/nvidia-persistenced/nvidia-persistenced.pid";
ExecStart = "${lib.getExe nvidia_x11.persistenced} --verbose";
ExecStopPost = "${pkgs.coreutils}/bin/rm -rf /var/run/nvidia-persistenced";
};
};
})
(lib.mkIf cfg.dynamicBoost.enable {
"nvidia-powerd" = {
description = "nvidia-powerd service";
path = [
pkgs.util-linux # nvidia-powerd wants lscpu
];
wantedBy = [ "multi-user.target" ];
serviceConfig = {
Type = "dbus";
BusName = "nvidia.powerd.server";
ExecStart = "${nvidia_x11.bin}/bin/nvidia-powerd";
};
};
})
];
services.acpid.enable = true;
services.dbus.packages = lib.optional cfg.dynamicBoost.enable nvidia_x11.bin;
hardware.firmware = lib.optional (cfg.open || lib.versionAtLeast nvidia_x11.version "555") nvidia_x11.firmware;
systemd.tmpfiles.rules =
[
# Remove the following log message:
# (WW) NVIDIA: Failed to bind sideband socket to
# (WW) NVIDIA: '/var/run/nvidia-xdriver-b4f69129' Permission denied
#
# https://bbs.archlinux.org/viewtopic.php?pid=1909115#p1909115
"d /run/nvidia-xdriver 0770 root users"
]
++ lib.optional (nvidia_x11.persistenced != null && config.virtualisation.docker.enableNvidia)
"L+ /run/nvidia-docker/extras/bin/nvidia-persistenced - - - - ${nvidia_x11.persistenced}/origBin/nvidia-persistenced";
boot = {
extraModulePackages = if cfg.open then [ nvidia_x11.open ] else [ nvidia_x11.bin ];
# nvidia-uvm is required by CUDA applications.
kernelModules = lib.optionals config.services.xserver.enable [
"nvidia"
"nvidia_modeset"
"nvidia_drm"
];
# If requested enable modesetting via kernel parameters.
kernelParams =
lib.optional (offloadCfg.enable || cfg.modesetting.enable) "nvidia-drm.modeset=1"
++ lib.optional ((offloadCfg.enable || cfg.modesetting.enable) && lib.versionAtLeast nvidia_x11.version "545") "nvidia-drm.fbdev=1"
++ lib.optional cfg.powerManagement.enable "nvidia.NVreg_PreserveVideoMemoryAllocations=1"
++ lib.optional cfg.open "nvidia.NVreg_OpenRmEnableUnsupportedGpus=1"
++ lib.optional (config.boot.kernelPackages.kernel.kernelAtLeast "6.2" && !ibtSupport) "ibt=off";
# enable finegrained power management
extraModprobeConfig = lib.optionalString cfg.powerManagement.finegrained ''
options nvidia "NVreg_DynamicPowerManagement=0x02"
'';
};
services.udev.extraRules = lib.optionalString cfg.powerManagement.finegrained (
lib.optionalString (lib.versionOlder config.boot.kernelPackages.kernel.version "5.5") ''
# Remove NVIDIA USB xHCI Host Controller devices, if present
ACTION=="add", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x0c0330", ATTR{remove}="1"
# Remove NVIDIA USB Type-C UCSI devices, if present
ACTION=="add", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x0c8000", ATTR{remove}="1"
# Remove NVIDIA Audio devices, if present
ACTION=="add", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x040300", ATTR{remove}="1"
''
+ ''
# Enable runtime PM for NVIDIA VGA/3D controller devices on driver bind
ACTION=="bind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030000", TEST=="power/control", ATTR{power/control}="auto"
ACTION=="bind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030200", TEST=="power/control", ATTR{power/control}="auto"
# Disable runtime PM for NVIDIA VGA/3D controller devices on driver unbind
ACTION=="unbind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030000", TEST=="power/control", ATTR{power/control}="on"
ACTION=="unbind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030200", TEST=="power/control", ATTR{power/control}="on"
''
);
})
# Data Center
(lib.mkIf (cfg.datacenter.enable) {
boot.extraModulePackages = [ nvidia_x11.bin ];
systemd = {
tmpfiles.rules =
lib.optional (nvidia_x11.persistenced != null && config.virtualisation.docker.enableNvidia)
"L+ /run/nvidia-docker/extras/bin/nvidia-persistenced - - - - ${nvidia_x11.persistenced}/origBin/nvidia-persistenced";
services = lib.mkMerge [
({
nvidia-fabricmanager = {
enable = true;
description = "Start NVIDIA NVLink Management";
wantedBy = [ "multi-user.target" ];
unitConfig.After = [ "network-online.target" ];
unitConfig.Requires = [ "network-online.target" ];
serviceConfig = {
Type = "forking";
TimeoutStartSec = 240;
ExecStart =
let
# Since these rely on the `nvidia_x11.fabricmanager` derivation, they're
# unsuitable to be mentioned in the configuration defaults, but they _can_
# be overridden in `cfg.datacenter.settings` if needed.
fabricManagerConfDefaults = {
TOPOLOGY_FILE_PATH = "${nvidia_x11.fabricmanager}/share/nvidia-fabricmanager/nvidia/nvswitch";
DATABASE_PATH = "${nvidia_x11.fabricmanager}/share/nvidia-fabricmanager/nvidia/nvswitch";
};
nv-fab-conf = settingsFormat.generate "fabricmanager.conf" (fabricManagerConfDefaults // cfg.datacenter.settings);
in
"${lib.getExe nvidia_x11.fabricmanager} -c ${nv-fab-conf}";
LimitCORE = "infinity";
};
};
})
(lib.mkIf cfg.nvidiaPersistenced {
"nvidia-persistenced" = {
description = "NVIDIA Persistence Daemon";
wantedBy = [ "multi-user.target" ];
serviceConfig = {
Type = "forking";
Restart = "always";
PIDFile = "/var/run/nvidia-persistenced/nvidia-persistenced.pid";
ExecStart = "${lib.getExe nvidia_x11.persistenced} --verbose";
ExecStopPost = "${pkgs.coreutils}/bin/rm -rf /var/run/nvidia-persistenced";
};
};
})
];
};
environment.systemPackages =
lib.optional cfg.datacenter.enable nvidia_x11.fabricmanager
++ lib.optional cfg.nvidiaPersistenced nvidia_x11.persistenced;
})
]
);
}
|