Jump to content
XPEnology Community

Nvidia Runtime Library

disone

By disone
in Synology Packages

 Share

Recommended Posts

disone Newbie

disone

Posted
Posted (edited)

Does anyone have any info on it? Can I use it with a nvidia GPU to accelerate my Plex?

source: https://archive.synology.com/download/Package/spk/NVIDIARuntimeLibrary/1.0.0-0011/

Requires DSM6.2-24833 on a synology_denverton_dva3219. But maybe someone ported it to the ds3615xs or other versions?

 

Edit:

I editted the INFO file in the package and it now will install on my DS3615xs @DSM 6.1 Will install a nvidia gpu to see what happens.

Edited by disone
Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted (edited)

there are no drivers in the package, its just x64 application code

denverton seems to use the same kernel as 918+ (4.4.59)

denvertons are:

1618+, 1819+, 2419+, dva3219, rs820+, rs820rp+, rs2418+, rs2418rp+, rs2818rp+

as it seems to be for dva3219 that one should at least contain the nvidia drivers (binary)  in the DSM_DVA3219_24922.pat

in /usr/lib/modules/

nvidia-modeset.ko

nvidia-uvm.ko

nvidia.ko

so as long as you dont have this files for 918+ the spk package will be of no use

edit: the nvidia files are also in the 1618+ so they might be in all denverton version

version numer in nvidia.ko is 381.22, its as old as the kernel synology used (~5/2017)

Edited by IG-88
  • Like 1
Link to comment
Share on other sites
  • 6 months later...
dragosmp Newbie

dragosmp

Posted
Posted
On 10/28/2019 at 11:30 PM, IG-88 said:

there are no drivers in the package, its just x64 application code

denverton seems to use the same kernel as 918+ (4.4.59)

denvertons are:

1618+, 1819+, 2419+, dva3219, rs820+, rs820rp+, rs2418+, rs2418rp+, rs2818rp+

as it seems to be for dva3219 that one should at least contain the nvidia drivers (binary)  in the DSM_DVA3219_24922.pat

in /usr/lib/modules/

nvidia-modeset.ko

nvidia-uvm.ko

nvidia.ko

so as long as you dont have this files for 918+ the spk package will be of no use

edit: the nvidia files are also in the 1618+ so they might be in all denverton version

version numer in nvidia.ko is 381.22, its as old as the kernel synology used (~5/2017)

Hi IG-88,

 

Is there a way that we can include these files in DSM6.2.3 for 918+ ? I have a Nvidia on a shelf and maybe I will do a better job than G5400 cpu.

  • Like 2
Link to comment
Share on other sites
  • 4 months later...
roysbike Newbie

roysbike

Posted
Posted
В 28.10.2019 в 22:56, disone сказал:

Does anyone have any info on it? Can I use it with a nvidia GPU to accelerate my Plex?

source: https://archive.synology.com/download/Package/spk/NVIDIARuntimeLibrary/1.0.0-0011/

Requires DSM6.2-24833 on a synology_denverton_dva3219. But maybe someone ported it to the ds3615xs or other versions?

 

Edit:

I editted the INFO file in the package and it now will install on my DS3615xs @DSM 6.1 Will install a nvidia gpu to see what happens.

hello! Send please file for ds3616xs

Link to comment
Share on other sites
  • 2 months later...
IG-88 Guru Master

IG-88

Posted
Posted (edited)

the driver for 3615 (NVIDIA-Linux-x86_64-440.44)

http://s000.tinyupload.com/?file_id=50915789462911185200

atm thats the only one i have as i was waiting to be able to test this later this month

3617 and 918+ will come later

 

the rest of the stuff that needs to be present (i guess) to work

https://archive.synology.com/download/Package/spk/NVIDIARuntimeLibrary/1.0.1-0014/

you would need to change the model entry of the INFO file in the spk (spk is just a zip and a changed INFO file can replace the original)

arch="x86_64"

and if there is a model="..." entry then remove that line

that way it should install in all x64 systems (including our 3 models)

 

i can't test this atm as i don't have matching card or a software that works without buying a license (like plex or emby) and it does not work for docker

 

Edited by IG-88
Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted
On 12/10/2020 at 3:12 AM, King Butter said:

Fresh install of Jun's loader 1.0.3b with DSM 6.23 emulating 3615xs

 

the message about what the unknown symbol are would have been good, that should be in the log (but i guess i found out about that)

anyway maybe try the older driver version i already used for a test the driver loading, it also need the older spk package from synology

 

nvidia drivers 381.22 (3615, 6.2.3) for using with NVIDIARuntimeLibrary-dva3219-1.0.0-0011.spk
(model in INFO needs to be removed to install on non dva System)

nvidia_381.22_3615_623.zip
http://www.filedropper.com/nvidia381223615623

 

copy the files to /usr/lib/modules

install the modded "NVIDIARuntimeLibrary-x86_64-1.0.0-0011.spk" (INFO in the spk needs to be modded)  

  

insmod /urs/lib/modules/nvidia.ko
insmod /urs/lib/modules/nvidia-uvm
insmod /urs/lib/modules/nvidia-drm
/usr/bin/nvidia-smi -pm 1

 

if  "nvidia-smi" is not present then extract it from the spk

/package.tgz/package/nvidia/bin/ and copy the files in bin to your /usr/bin folder

 

after that is loaded and and the spk installed it should be working but i was not able to verify this as it needed software that i did not have

Link to comment
Share on other sites
King Butter Newbie

King Butter

Posted
Posted (edited)

Sorry about that, I should have posted the error from dmesg. I tried a different approach, and it looks promising.

 

I went back to DS918+

Extracted nvidia modules from DS1618+ pat (there was no nvidia-drm module)

insmod /usr/lib/modules/nvidia.ko  

insmod /usr/lib/modules/nvidia-modeset.ko  

insmod /usr/lib/modules/nvidia-uvm.ko

Modified and installed NVIDIARuntimeLibrary-dva3221-1.0.1-0014.spk

/usr/local/bin/nvidia-smi gives me: 

Sat Dec 12 10:29:13 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.44       Driver Version: 440.44       CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 106...  On   | 00000000:01:00.0 Off |                  N/A |
| 32%   34C    P8     5W / 120W |      0MiB /  6078MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

 

 

/usr/local/bin/nvidia-smi -pm 1: 

Enabled persistence mode for GPU 00000000:01:00.0.
All done.

 

I'll work on testing with Plex tonight. 

 

I don't see a /dev/dri   is that normal for NVidia? is there something else I should be looking for?

Nevermind, I see it now

ls /dev/modules/nvid* 

/dev/nvidia0
/dev/nvidiactl

 

Edited by King Butter
  • Thanks 1
Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted (edited)
3 hours ago, King Butter said:

Extracted nvidia modules from DS1618+ pat (there was no nvidia-drm module)

interesting i did not know they where in that system, the only one i knew about where the DVA systems

how did you came to the conclusion to look there for nvidia drivers?, driver are usually only present if at least one device in that plattform has that hardware (like realtek driver in 918+ even when the 918+ does not use realtek nic)

edit: after looking it up, the DVA systems are denverton as the 1618+, that explains the presens of the nvidia drivers in 1618+ and the nvidia drivers of the dva *.pat should work in the same way as the 1618+ drivers for 918+

 

also i had no success in using drivers from another dsm system earlier (like 3617 <-> 3615), in that case the 1618+ is Denverton, Intel Atom C3538 vs Apollolake, Intel Celeron J3455

did you do anything else to the driver (like removing the signature)?

 

i guess the drm part might not be needed for transcoding, its present on "normal" linux system and loaded

if you look in /etc/rc, it does have a section about nvidia drivers and there a variable that can be set in synoinfo.conf to trigger the loading of the drivers when the system starts

supportnvgpu="yes"

 

 

3 hours ago, King Butter said:

/usr/local/bin/nvidia-smi -pm 1: 




Enabled persistence mode for GPU 00000000:01:00.0.
All done.

 

I'll work on testing with Plex tonight. 

 

 

nice, keep us updated if that construct works with plex, if driver and spk is enough then it should work with 3615/17 too (with *.ko files for nvidia)

 

Edited by IG-88
Link to comment
Share on other sites
King Butter Newbie

King Butter

Posted
Posted

Here's what I found in /etc/rc 

SYNOLoadNVGPUModule()
{
        local supportnvgpu=`get_key_value $SYNOINFO_DEF support_nvidia_gpu`

        if [ "$supportnvgpu" = "yes" ]; then
                SYNOLoadModules nvidia nvidia-uvm
                /usr/bin/nvidia-smi -pm 1
        fi
}

 

I thought to check for the drivers in the 1618 because of your previous post in the thread. 

 

Unfortunately, plex is not transcoding. My next step is to test with ffpmeg inside a docker container. I'll try that tonight or tomorrow and report back.

Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted
4 hours ago, King Butter said:

Unfortunately, plex is not transcoding. My next step is to test with ffpmeg inside a docker container. I'll try that tonight or tomorrow and report back.

there seemed to be things missing to have it working with docker (i wanted to try with jellyfin), i did not want to dive into trying to compile the whole nvidia package, just wanted to use the one already made by synology for dsm

 

ffmpeg is also a problem as it might not be aware of nvidiaenc for transcoding (that needs to be compiled into ffmpeg and it looked like the packages available (synology and 3rd party) where missing this feature (as, beside of the exotic dva units, nvidia drivers are not availible for dsm)

 

i thought plex might be a uiversal build for dsm, if the synology version is "optimized" not containing nvidia parts then it might be not working that way

 

Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted (edited)

 

maybe try a "older" plex version

 

"...

Linux-specific Notes

...

Starting with Plex Media Server v1.20.2, driver version 450.66 or newer is required for NVIDIA GPU usage.

..."

 

https://support.plex.tv/articles/115002178853-using-hardware-accelerated-streaming/

 

maybe this one?

https://downloads.plex.tv/plex-media-server-new/1.20.0.3181-0800642ec/synology/PlexMediaServer-1.20.0.3181-0800642ec-x86_64.spk

 

Edited by IG-88
Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted (edited)
18 hours ago, King Butter said:

Here's what I found in /etc/rc

about that, i looked in the synoinfo.conf of a dva unit and its 

support_nvidia_gpu="yes"

thats the line i overlooked

local supportnvgpu=`get_key_value $SYNOINFO_DEF support_nvidia_gpu`

here the yes from the synoinfo.conf variable is transferred to a new temporary one

Edited by IG-88
Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted
8 hours ago, King Butter said:

Still no HW transcoding

do you get something in /dev/dri from the nvidia drivers?

in theory there are different options to use transcoding, VAAPI (that would use /dev/dri devices) or NVENC

any log plex offers to see whats wrong with nvenc?

Link to comment
Share on other sites
King Butter Newbie

King Butter

Posted
Posted

I get /dev/nvidia0 and /dev/nvidiactl devices

 

Here's the relevant plex output: 

  

2616682:Dec 15, 2020 16:13:19.452 [0x7f455a840700] DEBUG - Codecs: testing h264_nvenc (encoder)
2616770:Dec 15, 2020 16:13:19.452 [0x7f455a840700] DEBUG - Codecs: hardware transcoding: testing API nvenc
2616869-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded lib: libcuda.so.1
2616956-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuInit
2617037-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDeviceGetCount
2617128-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDeviceGet
2617214-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDeviceGetAttribute
2617309-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDeviceGetName
2617399-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDeviceComputeCapability
2617499-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuCtxCreate_v2
2617588-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuCtxSetLimit
2617676-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuCtxPushCurrent_v2
2617770-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuCtxPopCurrent_v2
2617863-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuCtxDestroy_v2
2617953-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMemAlloc_v2
2618041-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMemAllocPitch_v2
2618134-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMemsetD8Async
2618224-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMemFree_v2
2618311-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMemcpy2D_v2
2618399-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMemcpy2DAsync_v2
2618492-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGetErrorName
2618581-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGetErrorString
2618672-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuStreamCreate
2618761-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuStreamQuery
2618849-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuStreamSynchronize
2618943-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuStreamDestroy_v2
2619036-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuStreamAddCallback
2619130-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuEventCreate
2619218-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuEventDestroy_v2
2619310-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuEventSynchronize
2619403-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuEventQuery
2619490-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuEventRecord
2619578-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuLaunchKernel
2619667-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuModuleLoadData
2619758-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuModuleUnload
2619847-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuModuleGetFunction
2619941-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuTexObjectCreate
2620033-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuTexObjectDestroy
2620126-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGLGetDevices_v2
2620218-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGraphicsGLRegisterImage
2620318-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGraphicsUnregisterResource
2620421-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGraphicsMapResources
2620518-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGraphicsUnmapResources
2620617-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuGraphicsSubResourceGetMappedArray
2620727-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDeviceGetUuid
2620817-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuImportExternalMemory
2620914-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDestroyExternalMemory
2621012-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuExternalMemoryGetMappedBuffer
2621118-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuExternalMemoryGetMappedMipmappedArray
2621232-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMipmappedArrayGetLevel
2621331-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuMipmappedArrayDestroy
2621429-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuImportExternalSemaphore
2621529-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuDestroyExternalSemaphore
2621630-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuSignalExternalSemaphoresAsync
2621736-Dec 15, 2020 16:13:19.453 [0x7f455a840700] DEBUG - [FFMPEG] - Loaded sym: cuWaitExternalSemaphoresAsync
2621840-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] - cu->cuInit(0) failed
2621923-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] -  -> CUDA_ERROR_UNKNOWN: unknown error
2622023-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] -
2622086-Dec 15, 2020 16:13:19.454 [0x7f455a840700] DEBUG - Codecs: hardware transcoding: opening hw device failed - probably not supported by this system, error: Unknown error occurred
2622263-Dec 15, 2020 16:13:19.454 [0x7f455a840700] DEBUG - MDE: Cannot direct stream video stream due to profile or setting limitations

 

Link to comment
Share on other sites
  • 3 weeks later...
robert36 Newbie

robert36

Posted
Posted

After reading log file that CUDA is not initialized:

 

2621840-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] - cu->cuInit(0) failed 2621923-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] - -> CUDA_ERROR_UNKNOWN: unknown error 2622023-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] - 

2621840-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] - cu->cuInit(0) failed
2621923-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] -  -> CUDA_ERROR_UNKNOWN: unknown error
2622023-Dec 15, 2020 16:13:19.454 [0x7f455a840700] ERROR - [FFMPEG] -

 

I found this web page:

 

https://forums.developer.nvidia.com/t/ffmpeg-cuda-fails/107144/8

 

The forum member have successfully run on linux custom compiled ffmpeg (we can take it from DS1618+ and need to try..if it is not working then it needs ffmpeg to be compiled for synology DS918+ using NVENC support and x.264 and x.265.

 

So above error means that linux driver is not installed completely as CUDA installation guide for linux on NVIDIA web page...and this why we get cuInit(0) FAILED error because we have not installed whole driver that can access to CUDA cores that is needed for transcoding...the fix is to read this page:

 

https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#verify-installation

 

And need to build and run  

./deviceQuery

and 

./vectorAdd

 

When above two commands return succes and show correct information about CUDA cores and Result = PASS ffmpeg will be working using NVIDIA NVENC transcoding...

 

For easy to spot what files we need to copy from DS1618+ ko drivers, i suggest to install on that PC where you have xpenology on second drive ubuntu and download these drivers for NVIDIA:

 

https://www.nvidia.com/Download/driverResults.aspx/164073/en-us

 

Install it on ubuntu, check ffmpeg transcoding and then we will know what files we need...then will be easy as nvidia in above package have .c and .h source code of driver, then we will know what driver we need to compile on DS918+ to match synology kernel version and x86_x64 arhitecture.

 

Compiling this way we can be always update with new NVIDIA driver and not to be waiting for synology to compile it for you. New drivers improves GPU performance and this way we can add support to any NVIDIA GPU that have pci expreee x16 and NVIDIA drivers and that have CUDA cores.

 

Waiting for someone to try this out...i don't have NVIDIA gpu so this is why i can't try...

Link to comment
Share on other sites
IG-88 Guru Master

IG-88

Posted
Posted (edited)
1 hour ago, robert36 said:

The forum member have successfully run on linux custom compiled ffmpeg (we can take it from DS1618+ and need to try..if it is not working then it needs ffmpeg to be compiled for synology DS918+ using NVENC support and x.264 and x.265.

there should be a proper ffmpeg version in the DVA3219 image or at least in the DVA version of the SurveillanceStation package (the dva version of SurveillanceStation uses nvidia features for tracking)

Edited by IG-88
Link to comment
Share on other sites
  • 2 weeks later...
lolo9419 Newbie

lolo9419

Posted
Posted (edited)

Hi,

 

I also added a quadro p400 in my Xpenolgy (3615xs and on a Hp microserver gen 8 ) and i'm interested of using it for Plex. After Doing some testing i was able to get at the same step as you, i get it to show my p400 with the command : 

/usr/local/bin/nvidia-smi 

but then it hangs forever and doesn't give me back the control... and if i start another ssh session and re-try /usr/local/bin/nvidia-smi.. it never does anything, just hang, only way to fix it is to restart. It does not crash the system at all, just create a dead process.
i do see the expected messages in dmesg 
  

[  359.424189] nvidia-nvlink: Nvlink Core is being initialized, major device number 247
[  359.424506] nvidia 0000:07:00.0: enabling device (0040 -> 0043)
[  359.424602] vgaarb: device changed decodes: PCI:0000:07:00.0,olddecodes=io+mem,decodes=none:owns=none
[  359.424699] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  381.22  Thu May  4 00:55:03 PDT 2017 (using threaded interrupts)
[  371.699376] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  381.22  Thu May  4 00:21:48 PDT 2017
[  379.116963] nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 246
[  415.006173] nvidia 0000:07:00.0: irq 62 for MSI/MSI-X
[  415.455850] 32bit 0000:07:00.0 uses non-identity mapping

and i get this from the hanged command :
  

[  601.188155] INFO: task nvidia-smi:24421 blocked for more than 120 seconds.
[  601.188322] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[  601.188505] nvidia-smi      D ffff8801e7a92040     0 24421  20685 0x00000000
[  601.188511]  ffff88011bc7bbf0 0000000000000082 000000000000c000 ffff88011bc7bfd8
[  601.188515]  ffff88011bc7bfd8 ffff88008a5f5040 0000000000000000 ffffffffa16ae37a
[  601.188518]  ffff880003688008 7fffffffffffffff 7fffffffffffffff ffff88011bc7bd28
[  601.188522] Call Trace:
[  601.188730]  [<ffffffffa16ae37a>] ? _nv012305rm+0x37a/0x400 [nvidia]
[  601.188737]  [<ffffffff814c0365>] ? schedule_timeout+0x155/0x1e0
[  601.188896]  [<ffffffffa18bed53>] ? _nv019507rm+0x23/0x70 [nvidia]
[  601.189051]  [<ffffffffa18beb07>] ? _nv019505rm+0x17/0x60 [nvidia]
[  601.189217]  [<ffffffffa18beb07>] ? _nv019505rm+0x17/0x60 [nvidia]
[  601.189228]  [<ffffffff814c1d60>] ? wait_for_completion+0x90/0xf0
[  601.189232]  [<ffffffff8105f110>] ? wake_up_state+0x10/0x10
[  601.189237]  [<ffffffff8104abf5>] ? flush_workqueue+0x115/0x5d0
[  601.189334]  [<ffffffffa13c29a2>] ? os_flush_work_queue+0x52/0x60 [nvidia]
[  601.189468]  [<ffffffffa198fab7>] ? rm_disable_adapter+0x77/0x130 [nvidia]
[  601.189562]  [<ffffffffa13b6837>] ? nv_close_device+0x77/0x170 [nvidia]
[  601.189657]  [<ffffffffa13ba8c7>] ? nvidia_close+0xd7/0x3a0 [nvidia]
[  601.189754]  [<ffffffffa13b6382>] ? nvidia_frontend_close+0x22/0x40 [nvidia]
[  601.189759]  [<ffffffff810fcf45>] ? __fput+0xc5/0x230
[  601.189763]  [<ffffffff810501ff>] ? task_work_run+0x8f/0xd0
[  601.189767]  [<ffffffff81002931>] ? do_notify_resume+0x61/0x80
[  601.189771]  [<ffffffff814c6067>] ? int_signal+0x12/0x17


I used these instructions/drivers to make it "works". As the more recent version of the driver given under or the one extracted from the pat of the DS1618+ (as suggested by King Butter) gives me the "Unknown symbol in module" when trying to inject into the kernel.

Le 12/12/2020 à 03:42, IG-88 a dit :

 

the message about what the unknown symbol are would have been good, that should be in the log (but i guess i found out about that)

anyway maybe try the older driver version i already used for a test the driver loading, it also need the older spk package from synology

 

nvidia drivers 381.22 (3615, 6.2.3) for using with NVIDIARuntimeLibrary-dva3219-1.0.0-0011.spk
(model in INFO needs to be removed to install on non dva System)

nvidia_381.22_3615_623.zip
http://www.filedropper.com/nvidia381223615623

 

copy the files to /usr/lib/modules

install the modded "NVIDIARuntimeLibrary-x86_64-1.0.0-0011.spk" (INFO in the spk needs to be modded)  

  


insmod /urs/lib/modules/nvidia.ko
insmod /urs/lib/modules/nvidia-uvm
insmod /urs/lib/modules/nvidia-drm
/usr/bin/nvidia-smi -pm 1

 

if  "nvidia-smi" is not present then extract it from the spk

/package.tgz/package/nvidia/bin/ and copy the files in bin to your /usr/bin folder

 

after that is loaded and and the spk installed it should be working but i was not able to verify this as it needed software that i did not have

Trying on plex.. It's also detected but i get the same error  : 

[Transcode] [FFMPEG] - -> CUDA_ERROR_UNKNOWN: unknown error

 

I'm probably not helping you to make any progress, but i thought i would comment what i did in case it might help .

Thanks to everyone for the work trying to figure this out. If i can help in any way, let me know ! :)

Edited by lolo9419
Link to comment
Share on other sites
RedwinX Super Member

RedwinX

Posted
Posted

Hey All, 

 

Very intresting topic. I have installed a GT 1030 in a microserver gen 8, with DSM 6.2.3. I have compiled a custom version of ffmpeg, and still have the same error, cuda initialization failed. 

Driver is ok (I think) :

image.png.aa8215bdd5ce636cc4ba4bfed6d158da.png

image.png.23f1d62003afe1c3b35ca2773e656aa9.png

 

Trying to install NVIDIA Driver : 

image.png.defbcb2b91a0f43848ab442d2d7cc205.png

  • Thanks 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
 Share
Go to topic listing
×
×
  • Create New...