DaVinci Resolve on Linux - Install issues

@Dwaine: Thanks for testing. My settings were already in "Debayer only" mode in order to use more CPU cores. I've changed to "Debayer+Decompression" and no crash yet. I have to try more and more because it occurs very randomly. By the way, if I select "None", RED files appear as offline as you already know. That's something new since version 17.

I've had a case of if it ain't broke, don't fix it. I was on v17.0 and everything was working fine, but I decided to upgrade to v17.2.2 today and everything seem to still work fine, except for one thing. Now my braw files aren't showing up in the media tab anymore. Other formats like mov and mp4 shows up, just not braw. A bit ironic I must say.

Has anyone else experienced this?

I installed by simply doing

Code: Select all

sudo ./DaVinci_Resolve_Studio_17.2.2_Linux.run -i

as that has worked the last couple of times. Not sure if installing with makeresolvedeb will make any difference. The braw files comes from the bmccp6k running v6.9.6.

Computer info:
OS: Pop!_OS 21.04 x86_64
Kernel: 5.11.0-7620-generic
GPU: NVIDIA GeForce GTX 1070 Mobile (v470.57.02)

The first time I opened resolve I got a welcome screen that was just blank, once I force closed it resolve launched and it hasn't come up again. Probably has nothing to do with it, I just thought I'd mention it in case.

Any ideas how to solve this?

EDIT:
No worries, I just tried the makeresolvedeb method and now the braw files appear

I'm trying out 3 laptops with Ubuntu 20.04 and Resolve 17.2.2:

GIGABYTE - 15.6" 4K OLED Creator Laptop - Intel Core i7-11800H - 16GB - NVIDIA GeForce RTX 3060 1TB SSD

HP OMEN - 15.6" Gaming Laptop - AMD Ryzen 7 - 16GB Memory - NVIDIA GeForce RTX 3060 - 512GB SSD

ASUS - ROG Zephyrus 14" Gaming Laptop - AMD Ryzen 9 - 16GB Memory - NVIDIA GeForce RTX 3060 - 1TB SSD

The ASUS works extremely well. Better than my desktops. The other two pop the old GPU Error Code 999.

But the ASUS is just 14" and hard to see. The problem with the other two seems to be:

"DVIP Exception: At position /home/jenkins/jenkins/workspace/dvip/DaVinciIP/DaVinciIP/Debayer/Debayer_kernel.cu:318"

Anyone have any ideas before I start returns?

I'm using Mint 20.2 and nVidia proprietary drivers (470.57.02 - also tried 460.91.03) which always worked fine until now.
Having pocl packages installed, resolve runs into an segmentation fault right after start.
After removing the pocl packages it works again.

Here's the console output:

Code: Select all

 /opt/resolve/bin/resolve
ActCCMessage Already in Table: Code= c005, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c006, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c007, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= 2282, Mode=  0, Level=  0, CmdKey= 8, Option= 0
PnlMsgActionStringAdapter Already in Table: Code= 615e, Mode=  0, Level=  0, CmdKey= -1, Option= 0
17.2.2.0004 Linux/Clang x86_64
Main thread starts: 21B77680
0x7f5121b77680 | Undefined            | INFO  | 2021-08-15 12:56:10,749 | --------------------------------------------------------------------------------
0x7f5121b77680 | Undefined            | INFO  | 2021-08-15 12:56:10,749 | Loaded log config from /home/x/.local/share/DaVinciResolve/configs/log-conf.xml
0x7f5121b77680 | Undefined            | INFO  | 2021-08-15 12:56:10,749 | --------------------------------------------------------------------------------


gdb backtrace:

Code: Select all

Thread 1 "resolve" received signal SIGSEGV, Segmentation fault.
b0x00007fffc5d73660 in ?? ()
(gdb) bt
#0  0x00007fffc5d73660 in  ()
#1  0x00000000080f8959 in std::error_code::default_error_condition() const ()
#2  0x00007fffbb90e175 in  () at /lib/x86_64-linux-gnu/libpocl.so.2.4.0
#3  0x00007fffbb8ed7bf in  () at /lib/x86_64-linux-gnu/libpocl.so.2.4.0
#4  0x00007fffbb8ee7e3 in  () at /lib/x86_64-linux-gnu/libpocl.so.2.4.0
#5  0x00007fffbb8cb2ab in  () at /lib/x86_64-linux-gnu/libpocl.so.2.4.0
#6  0x00007fffea730b17 in  () at /lib/x86_64-linux-gnu/libOpenCL.so.1
#7  0x00007fffea7310d3 in clGetPlatformIDs () at /lib/x86_64-linux-gnu/libOpenCL.so.1
#8  0x00007ffff6ac3ff6 in  () at /media/data/system/resolve/bin/../libs/libgpudetect.so
#9  0x00007ffff6ae61e0 in  () at /media/data/system/resolve/bin/../libs/libgpudetect.so
#10 0x00007ffff6ae5b15 in  () at /media/data/system/resolve/bin/../libs/libgpudetect.so
#11 0x000000000262aca2 in  ()
#12 0x00000000026276ab in  ()
#13 0x00007fffdab910b3 in __libc_start_main (main=0x2626df0, argc=1, argv=0x7fffffffde18, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffffffde08)
    at ../csu/libc-start.c:308
#14 0x00000000026262bb in  ()


[Dwaine Maggart]

17.3 now has ALSA audio support. ADR is now available.

17.3 now has ALSA audio support. ADR is now available.

I can confirm ALSA record/ADR now working. This is great! Thank you!

No worries, I just tried the makeresolvedeb method and now the braw files appear

Yes, makeresolvedeb contains fixes to make the braw lib load properly.

The first time I opened resolve I got a welcome screen that was just blank, once I force closed it resolve launched and it hasn't come up again. Probably has nothing to do with it, I just thought I'd mention it in case.

The blank welcome screen is probably due to incompatible OS libraries (my guess is libssl). A quite common error. You can omit the onboarding entirely by supplying the argument --skip-onboarding to makeresolvedeb. It doesn't fix the onboarding but at least Resolve will start properly.

Problems with CentOS and AMD Ellsmore (RX570) card?

Should I just assume that if you have an AMD card, running Resolve on Linux (and not using an nVidia card) is darn near impossible?

I've searched and searched and I haven't found anything that would help me with this error when I attempt to start Resolve: "....error while loading shared libraries: libOpenCL.so.1: cannot open shared object file: No such file or directory..."

I'm on CentOS 7.9 (7.3 spit out the same error), my computer is an Asus Z370 with a 9700K, and the RX570 (4GB). I also tried the official AMD drivers, and same result.

I tried the yum install and can't find the equivalent of apt get install for that library.

Any thoughts out there?

Thanks in advance.

I can confirm ALSA record/ADR now working. This is great! Thank you!

Works here too! Now trying to get it to work with my X-Air 18.

EDIT: Spoke a bit too soon. Yes, Resolve is recording audio, but I'm getting crackling on both machines I've tried that use Behringer USB mixers as input (X-Air 18 and Q802USB). Both work fine with Audacity and other programs. Audio playback does not follow the timeline.

OK, I gave the AMD amdgpu-pro-20.45-1188099-rhel-8.2.tar driver another shot.
I used this install argument:
./amdgpu-pro-install -y --opencl=rocr,legacy
That seemed to work.
So both Resolve 16.2.7 and 17b9 Studio work with the RX580.
You do get H.264 and H.265 playback. 8 and 10 bit. HD and UHD.
You get no H.264/H.265 encode options. No H.264/5 under QuickTime and no MP4 selection.

Dwaine: I was able to get Resolve started on my Z370 with this AMD install command (Resolve 17.2.2). I haven't worked much with it, but it seems fine, with one caveat. Will know more later about how well it works, but about that caveat......

Well, here it is: On CentOS, I cannot get rid of the nomodeset command. I removed it from the grub.cfg (/etc/default/grub), and ran grub-mkconfig several times, but it always comes back. I can guarantee it is not in /etc/default/grub. How did you get rid of it?

Because, the way I get in now is I hold down the "e" key and remove the line that references nomodeset, and then the system boots. If I don't do that, the system does not boot.

How did you get rid of it?

Thanks in advance!

[Dwaine Maggart]

I didn't have that issue, so I'm afraid I don't have an answer.

I didn't have that issue, so I'm afraid I don't have an answer.

It's ok, I finally figured it out. You have to put in something for nomodeset and I put in "quiet splash" and then re-did the rebuild of grub2. CentOS is kind of difficult as it's not widely used, there's not a lot of support out there (unlike ubuntu, etc) and it also doesn't have many of the utilities I'm used to.

Thanks anyway.

P.S. - The good news! Resolve works on this linux install (and with an AMD card!)

I can confirm ALSA record/ADR now working. This is great! Thank you!

Works here too! Now trying to get it to work with my X-Air 18.

EDIT: Spoke a bit too soon. Yes, Resolve is recording audio, but I'm getting crackling on both machines I've tried that use Behringer USB mixers as input (X-Air 18 and Q802USB). Both work fine with Audacity and other programs. Audio playback does not follow the timeline.

Yes, I also get the crackling. I have tried a couple of different USB audio interfaces. Also don't have that problem with other audio apps.

I can confirm ALSA record/ADR now working. This is great! Thank you!

Works here too! Now trying to get it to work with my X-Air 18.

EDIT: Spoke a bit too soon. Yes, Resolve is recording audio, but I'm getting crackling on both machines I've tried that use Behringer USB mixers as input (X-Air 18 and Q802USB). Both work fine with Audacity and other programs. Audio playback does not follow the timeline.

Yes, I also get the crackling. I have tried a couple of different USB audio interfaces. Also don't have that problem with other audio apps.

Are you running pulseaudio by any chance? I've had sync and crackle problems in several applications including Resolve even before version 17. Samplerate conversion in pulseaudio is a little temperamental in my experience. I fixed it by setting default-sample-rate = 48000 in /etc/pulse/daemon.conf and keeping all source and sink devices at 48000. I already had this set on my Resolve workstation before 17.3 so I can't say if that is the solution but I've had no problem so far (knock och wood). I'm no big fan of pulseaudio but it's what we are stuck with now on most distros, and don't get me started on it's partner in crime; systemd...

I fixed it by setting default-sample-rate = 48000 in /etc/pulse/daemon.conf and keeping all source and sink devices at 48000.

I tried that, but unfortunately I think it's more complicated.

With a mic connected to input 1, Resolve routes input to channel 2, i.e., they are reversed. But on output, they are correct.

During record, random pops appear and are recorded on the stereo channel with no input. Using a mono channel does not help.

I have duplicated this on two machines (Xeon/21.04 and Threadripper/20.04), and then swapped USB sources (X-Air 18 and Q802USB), all with the same result.

I'm curious if anyone has had any issue with Volume disconnect issues? Sometimes Resolve will work fine and suddenly, it will completely disconnect me from my internal RAID.

Specifically, every time I turn on Render Caching, I lose a connection to my media. When I try to connect to my Storage via linux, it is unmounted. I takes a restart to get it back.

***update***
Looks like this is just a good old permissions issue. I was being locked out of the storage mount. I created a new folder for the cache and all is good now.

***XTRA update***
Nope. Still having issues. Should mention this RAID is shared on the network via NFS. Still searching for a solution.

Problems with CentOS and AMD Ellsmore (RX570) card?

Should I just assume that if you have an AMD card, running Resolve on Linux (and not using an nVidia card) is darn near impossible?

I've searched and searched and I haven't found anything that would help me with this error when I attempt to start Resolve: "....error while loading shared libraries: libOpenCL.so.1: cannot open shared object file: No such file or directory..."

I'm on CentOS 7.9 (7.3 spit out the same error), my computer is an Asus Z370 with a 9700K, and the RX570 (4GB). I also tried the official AMD drivers, and same result.

I tried the yum install and can't find the equivalent of apt get install for that library.

Any thoughts out there?

Thanks in advance.

No you shouldn't assume that at all. There are oodles of people running RX580s including myself. I used to run it on Centos 7 but I am running on 8 now. The thing is you can't use the opensource openCL drivers and you can't use ROCm. The solution for me was to use the AMDGPU-pro drivers.

If you back pedal a few pages some one wrote a script to pull out just certain components so you can use your distros mesa for example with the proprietary openCL drivers.

In my case I have fedora as my main desktop with the distro drivers then dual boot to Centos where I have the proprietary drivers., That OS never gets patched ever, because yup AMDs drivers couldn't suck worse. Bring on competition in the form of Intel.

I did encounter segfaults when starting resolve on Fedora 33 after upgrading to the latest 17.3. An upgrade to F34 didn't help either (with the latest 470.63.01 nvidia driver via rpmfusion).
Looking at `dmesg` revealed the following:

Code: Select all

[  117.722198] NVRM: API mismatch: the client has the version 440.33.01, but
               NVRM: this kernel module has the version 470.63.01.  Please
               NVRM: make sure that this kernel module and all NVIDIA driver
               NVRM: components have the same version.


I had leftovers from previous non-rpm based nvidia driver installations still in the library path:

Code: Select all

-rwxr-xr-x. 1 root root  1285296 Nov 13  2019 libEGL_nvidia.so.440.33.01
-rwxr-xr-x. 1 root root    63696 Nov 13  2019 libGLESv1_CM_nvidia.so.440.33.01
-rwxr-xr-x. 1 root root   111416 Nov 13  2019 libGLESv2_nvidia.so.440.33.01
-rwxr-xr-x. 1 root root  1114496 Nov 13  2019 libGLX_nvidia.so.440.33.01
-rwxr-xr-x. 1 root root 16591136 Nov 13  2019 libcuda.so.440.33.01
-rwxr-xr-x. 1 root root 16711200 Feb 26  2020 libcuda.so.440.64.00
-rwxr-xr-x. 1 root root  3218784 Nov 13  2019 libnvcuvid.so.440.33.01
-rwxr-xr-x. 1 root root  3218880 Feb 26  2020 libnvcuvid.so.440.64.00
-rwxr-xr-x. 1 root root    78080 Nov 13  2019 libnvidia-allocator.so.440.33.01
-rwxr-xr-x. 1 root root   402136 Nov 13  2019 libnvidia-cbl.so.440.33.01
-rwxr-xr-x. 1 root root   200944 Nov 13  2019 libnvidia-cfg.so.440.33.01
-rwxr-xr-x. 1 root root 49976536 Nov 13  2019 libnvidia-compiler.so.440.33.01
-rwxr-xr-x. 1 root root 49976536 Feb 26  2020 libnvidia-compiler.so.440.64.00
-rwxr-xr-x. 1 root root 26747264 Nov 13  2019 libnvidia-eglcore.so.440.33.01
-rwxr-xr-x. 1 root root    99536 Nov 13  2019 libnvidia-encode.so.440.33.01
-rwxr-xr-x. 1 root root   103632 Feb 26  2020 libnvidia-encode.so.440.64.00
-rwxr-xr-x. 1 root root   386168 Nov 13  2019 libnvidia-fatbinaryloader.so.440.33.01
-rwxr-xr-x. 1 root root   135560 Nov 13  2019 libnvidia-fbc.so.440.33.01
-rwxr-xr-x. 1 root root 28623336 Nov 13  2019 libnvidia-glcore.so.440.33.01
-rwxr-xr-x. 1 root root   683112 Nov 13  2019 libnvidia-glsi.so.440.33.01
-rwxr-xr-x. 1 root root 11114264 Nov 13  2019 libnvidia-glvkspirv.so.440.33.01
-rw-r--r--. 1 root root  5227432 Dec 11  2019 libnvidia-gtk3.so.440.44
-rwxr-xr-x. 1 root root   207064 Nov 13  2019 libnvidia-ifr.so.440.33.01
-rwxr-xr-x. 1 root root  1465752 Nov 13  2019 libnvidia-ml.so.440.33.01
-rwxr-xr-x. 1 root root  1465880 Feb 26  2020 libnvidia-ml.so.440.64.00
-rwxr-xr-x. 1 root root 29063816 Nov 13  2019 libnvidia-opencl.so.440.33.01
-rwxr-xr-x. 1 root root 29182184 Feb 26  2020 libnvidia-opencl.so.440.64.00
-rwxr-xr-x. 1 root root    40720 Nov 13  2019 libnvidia-opticalflow.so.440.33.01
-rwxr-xr-x. 1 root root    40720 Feb 26  2020 libnvidia-opticalflow.so.440.64.00
-rwxr-xr-x. 1 root root  8873656 Nov 13  2019 libnvidia-ptxjitcompiler.so.440.33.01
-rwxr-xr-x. 1 root root  8873656 Feb 26  2020 libnvidia-ptxjitcompiler.so.440.64.00
-rwxr-xr-x. 1 root root 32987072 Nov 13  2019 libnvidia-rtcore.so.440.33.01
-rwxr-xr-x. 1 root root    14480 Nov 13  2019 libnvidia-tls.so.440.33.01
-rwxr-xr-x. 1 root root 93996912 Nov 13  2019 libnvoptix.so.440.33.01


That was apparently not an issue with older versions. After removing those old libraries, resolve starts fine without any segfaults.

Any idea why loop can refuse to work on any page?
Alt+/ keys
Works on windows. v17.3.1 thanks!

I'm curious if anyone has had any issue with Volume disconnect issues? Sometimes Resolve will work fine and suddenly, it will completely disconnect me from my internal RAID.

Specifically, every time I turn on Render Caching, I lose a connection to my media. When I try to connect to my Storage via linux, it is unmounted. I takes a restart to get it back.

***update***
Looks like this is just a good old permissions issue. I was being locked out of the storage mount. I created a new folder for the cache and all is good now.

***XTRA update***
Nope. Still having issues. Should mention this RAID is shared on the network via NFS. Still searching for a solution.

So, I was able to explore a little further at what seems to be causing an issue. If Resolve creates a folder on my RAID, I will get locked out of the RAID. If I render to folders created by another user, everything is fine. The main issue is when render caching is turned on, folders get created, and then I get shut out.

Any ideas how to fix?

I'm trying to install Resolve on Ubuntu 20 - I built a deb file as directed, and I get a configuration menu starting, but once that has been completed, the main program totally fails to load. There isn't much in the logs to go on.

Code: Select all

richja@beasty:/opt/resolve$ ./bin/resolve
ActCCMessage Already in Table: Code= c005, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c006, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c007, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= 2282, Mode=  0, Level=  0, CmdKey= 8, Option= 0
PnlMsgActionStringAdapter Already in Table: Code= 615e, Mode=  0, Level=  0, CmdKey= -1, Option= 0
17.3.1.0005 Linux/Clang x86_64
Main thread starts: 82C75680
0x7f2582c75680 | Undefined            | INFO  | 2021-09-03 20:40:39,776 | --------------------------------------------------------------------------------
0x7f2582c75680 | Undefined            | INFO  | 2021-09-03 20:40:39,776 | Loaded log config from /home/richja/.local/share/DaVinciResolve/configs/log-conf.xml
0x7f2582c75680 | Undefined            | INFO  | 2021-09-03 20:40:39,776 | --------------------------------------------------------------------------------
Segmentation fault (core dumped)
richja@beasty:/opt/resolve$ cat /home/richja/.local/share/DaVinciResolve/logs/
LogArchive/       ResolveDebug.txt 
richja@beasty:/opt/resolve$ cat /home/richja/.local/share/DaVinciResolve/logs/ResolveDebug.txt
0x7f2582c75680 | Main                 | INFO  | 2021-09-03 20:40:39,776 | Running DaVinci Resolve v17.3.1.0005 (Linux/Clang x86_64)
0x7f2582c75680 | Main                 | INFO  | 2021-09-03 20:40:39,776 | BMD_BUILD_UUID 0613bd84-384e-4843-8b9a-c9580b9c44e4
0x7f2582c75680 | Main                 | INFO  | 2021-09-03 20:40:39,776 | BMD_GIT_COMMIT 0794d58e3a0a02d22ebea304b92345b82ab7e5af
0x7f2582c75680 | GPUDetect            | INFO  | 2021-09-03 20:40:39,777 | Starting GPUDetect 1.1_3-a5


ldd on the binary shows all the libraries are found, the system checker indicated that the GPU device was OK.

Any suggestions very welcomed. Thank you,
James

I'm trying to install Resolve on Ubuntu 20 - I built a deb file as directed, and I get a configuration menu starting, but once that has been completed, the main program totally fails to load. There isn't much in the logs to go on.

Code: Select all

richja@beasty:/opt/resolve$ ./bin/resolve
ActCCMessage Already in Table: Code= c005, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c006, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c007, Mode= 13, Level=  1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= 2282, Mode=  0, Level=  0, CmdKey= 8, Option= 0
PnlMsgActionStringAdapter Already in Table: Code= 615e, Mode=  0, Level=  0, CmdKey= -1, Option= 0
17.3.1.0005 Linux/Clang x86_64
Main thread starts: 82C75680
0x7f2582c75680 | Undefined            | INFO  | 2021-09-03 20:40:39,776 | --------------------------------------------------------------------------------
0x7f2582c75680 | Undefined            | INFO  | 2021-09-03 20:40:39,776 | Loaded log config from /home/richja/.local/share/DaVinciResolve/configs/log-conf.xml
0x7f2582c75680 | Undefined            | INFO  | 2021-09-03 20:40:39,776 | --------------------------------------------------------------------------------
Segmentation fault (core dumped)
richja@beasty:/opt/resolve$ cat /home/richja/.local/share/DaVinciResolve/logs/
LogArchive/       ResolveDebug.txt 
richja@beasty:/opt/resolve$ cat /home/richja/.local/share/DaVinciResolve/logs/ResolveDebug.txt
0x7f2582c75680 | Main                 | INFO  | 2021-09-03 20:40:39,776 | Running DaVinci Resolve v17.3.1.0005 (Linux/Clang x86_64)
0x7f2582c75680 | Main                 | INFO  | 2021-09-03 20:40:39,776 | BMD_BUILD_UUID 0613bd84-384e-4843-8b9a-c9580b9c44e4
0x7f2582c75680 | Main                 | INFO  | 2021-09-03 20:40:39,776 | BMD_GIT_COMMIT 0794d58e3a0a02d22ebea304b92345b82ab7e5af
0x7f2582c75680 | GPUDetect            | INFO  | 2021-09-03 20:40:39,777 | Starting GPUDetect 1.1_3-a5


ldd on the binary shows all the libraries are found, the system checker indicated that the GPU device was OK.

Any suggestions very welcomed. Thank you,
James

What GPU are you using and which driver version?

Hard Drive Crash

And just when I got the linux install done and the AMD drivers installed and working, I had to rebuild everything.

Couldn't get Resolve CentOS drive to boot and install, so I had to go with a vanilla CentOS 7.9.

Same I had before with the AMD drivers, I can't get the AMD drivers to install, but a different error this time:

sudo ./amdgpu-pro-install -y --opencl=rocr,legacy

Loaded plugins: fastestmirror, langpacks
Loading mirror speeds from cached hostfile
* base: mirrors.raystedman.org
* elrepo: dfw.mirror.rackspace.com
* epel: mirror.prgmr.com
* extras: mirror.san.fastserv.com
* nux-dextop: mirror.li.nux.ro
* updates: mirrors.ocf.berkeley.edu
No package amdgpu-pro-versionlist available.
Error: Nothing to do

Anybody out there got any ideas as to what it's looking at or what is missing?

[Dwaine Maggart]

Did you see my earlier post relating to AMD on CentOS 7.9?

Haven't tried since then, but this worked for me:

viewtopic.php?f=21&t=56878&p=727436&hilit=amd#p727436

Actually, I see you used my install arguments. Did you use the amdgpu-pro-20.45-1188099-rhel-8.2.tar driver?

Did you see my earlier post relating to AMD on CentOS 7.9?

Haven't tried since then, but this worked for me:

viewtopic.php?f=21&t=56878&p=727436&hilit=amd#p727436

Actually, I see you used my install arguments. Did you use the amdgpu-pro-20.45-1188099-rhel-8.2.tar driver?

Yes, that is exactly what I used. I just don't understand the error in light of that I'm using CentOS, with the same AMD driver and the same version of Resolve as I did before. Perhaps the Blackmagic CentOS has some adjustments that the vanilla version doesn't have, but I couldn't geet the USB stick to install. It would boot, but would freeze shortly after it scanned the drives, and it wouldn't continue. I tried several different USB drives, but gave up after a while, and just went with the vanilla CentOS 7.9 install.

Don't know what's different or not right because I can't get Resolve to start without that AMD driver.

[Dwaine Maggart]

Our CentOS 7.3 ISO will only install from a DVD. And it will only work with an NVIDIA GPU. And will not work if there are any active onboard graphics.

I would suggest anyone building a Resolve Linux system these days to use the Seth Goldin Resolve CentOS install process:

https://github.com/sethgoldin/install-d ... lve-centos

That will install from a USB stick and is not sensitive to onboard graphics.

He has a process for CentOS 7.9 and 8.3.

And his process should work for AMD builds as well, if you replace the NVIDIA install steps with whatever the heck the AMD driver takes to install.

FIXED!!!

So, just for others, or anyone who needs to follow along, this is what I did:

1. Downloaded and Installed CentOS 7.3 (a generic version, not the Blackmagic ISO version) with a USB boot drive that I created with Rufus

2. Then I updated to CentOS 7.9 (sudo yum install updates - this got me to 7.9)

3. Followed the steps that were outlined here in the post that follows (which was very, very helpful, but as I have an AMD graphics card, I didn't install the nVidia drivers:

https://github.com/sethgoldin/install-d ... lve-centos
4. Because I was on 7.9, the drivers earlier referenced would not work (because they were 8.2 drivers! duh.....)
This driver worked: amdgpu-pro-20.45-1188099-rhel-7.9

5. Then I used Dwaine's magic install command line for the AMD drivers
-> sudo ./amdgpu-pro-install -y –opencl=rocr,legacy

And as soon as I was done, Resolve started right up!

6. Immediately, I cloned my new install (because I lost my previous due to disk failure), and then all done.

Now I need to resolve some issues with the Quicktime wrapper, but that's not for this forum......

I had a terrible issue when installing version 17.4 on Centos 8.2. When loading a project at first start, it crashed completely. After a forced reboot without the GUI (terminal), I could see a message regarding ALSA and Blackmagic Desktop driver (latest version by the way). Unfortunately, I didn't write it... I tried to uninstall the Desktop driver on the way from the terminal. No success (maybe I miss the right command line). After a second reboot, I got another error regarding the system: "failed to start switch root".

I had to restore a backup from the complete disk

EDIT : Back to my previous install with an ALSA check status gives this result:

Code: Select all

aplay -l
**** List of PLAYBACK Hardware Devices ****
card 0: BlackmagicIO [BlackmagicIO Audio], device 0: PCM [PCM]
  Subdevices: 1/1
  Subdevice #0: subdevice #0
card 1: Audio [USB Audio], device 0: USB Audio [USB Audio]
  Subdevices: 0/1
  Subdevice #0: subdevice #0

@Lucius: Testing your project on an HP z420 with CentOS 8.3 and 17.2.2 and an RTX6000 GPU (closest thing I have to your RTX3090) I'm not seeing any issues. I played with it randomly yesterday over the day, and let it play overnight. That was with no local media. Today, I've added an 8K Weapon clip and applied your clip 2 grades to it. And no issues.

Maybe it's related to your Komodo footage. I don't have any of that to test with.

If you have Red GPU Decompress and Debayer selected, try just selecting Debayer, and see if that has any effect.

Hi Dwain,

The problem has happened again today. No RED sample this time. Just only one DNxHR file on the timeline. The freezes occurs randomly, like in my previous post in July, and only on HDR project with Dolby Vision enabled.

One thing we've noticed: juste before the software freezing, the Mini Panel freezes too (plug by PoE). When you did some tests on your Centos machine, did you have the Panel plug in?

Thank you.

Hi,

I have upgraded Centos 8, nVidia drivers ans then DaVinci studio.

Now it crashes at startup:

Code: Select all

==========[CRASH DUMP]==========
#TIME Mon Dec 27 15:26:08 2021 - Uptime 00:00:00 (hh:mm:ss)
#PROGRAM_NAME DaVinci Resolve Studio v17.4.3.0010 (Linux/Clang x86_64)
#BMD_ARCHITECTURE x86_64
#BMD_BUILD_UUID c7e52229-4c8e-4499-9764-d355b026ac4e
#BMD_GIT_COMMIT 8d3b6611302940a18771c32043157e6fc072e38e
#BMD_UTIL_VERSION 17.4.3.0010
#OS Linux

/opt/resolve/bin/resolve() [0x4124399]
/opt/resolve/bin/resolve() [0x2148362]
/lib64/libpthread.so.0(+0x12c20) [0x7fd1a3fe1c20]
/lib64/libc.so.6(gsignal+0x10f) [0x7fd187bc137f]
/lib64/libc.so.6(abort+0x127) [0x7fd187babdb5]
/lib64/libc.so.6(+0x21c89) [0x7fd187babc89]
/lib64/libc.so.6(+0x2fa76) [0x7fd187bb9a76]
/opt/resolve/bin/resolve() [0x21718b6]
/opt/resolve/bin/resolve() [0x216fa78]
/opt/resolve/bin/resolve() [0x216f388]
/opt/resolve/bin/resolve() [0x3af00c8]
/opt/resolve/bin/resolve() [0x35a3e7d]
/opt/resolve/bin/resolve() [0x35a3d0e]
/opt/resolve/bin/resolve() [0x218e966]
/opt/resolve/bin/resolve() [0x218e8e0]
/opt/resolve/bin/resolve() [0x214658d]
/opt/resolve/bin/resolve() [0x21424bf]
/lib64/libc.so.6(__libc_start_main+0xf3) [0x7fd187bad493]
/opt/resolve/bin/resolve() [0x214153b]
Signal Number = 6

================================
resolve: /home/jenkins/jenkins/workspace/resolve/Resolve/Cyclone/Main/AppConfig.cpp:204: void AppConfig::LoadAllSiteInfo(): Assertion `m_SiteEnabledIdx > 0' failed.


==========[CRASH DUMP]==========
#TIME Mon Dec 27 15:26:08 2021 - Uptime 00:00:00 (hh:mm:ss)
#PROGRAM_NAME DaVinci Resolve Studio v17.4.3.0010 (Linux/Clang x86_64)
#BMD_ARCHITECTURE x86_64
#BMD_BUILD_UUID c7e52229-4c8e-4499-9764-d355b026ac4e
#BMD_GIT_COMMIT 8d3b6611302940a18771c32043157e6fc072e38e
#BMD_UTIL_VERSION 17.4.3.0010
#OS Linux

/opt/resolve/bin/resolve() [0x4124399]
/opt/resolve/bin/resolve() [0x2148362]
/lib64/libpthread.so.0(+0x12c20) [0x7fd1a3fe1c20]
/lib64/libc.so.6(gsignal+0x10f) [0x7fd187bc137f]
/lib64/libc.so.6(abort+0x127) [0x7fd187babdb5]
/lib64/libc.so.6(+0x21c89) [0x7fd187babc89]
/lib64/libc.so.6(+0x2fa76) [0x7fd187bb9a76]
/opt/resolve/bin/resolve() [0x21718b6]
/opt/resolve/bin/resolve() [0x216fa78]
/opt/resolve/bin/resolve() [0x216f388]
/opt/resolve/bin/resolve() [0x2180d1b]
/opt/resolve/bin/resolve() [0x2148551]
/lib64/libpthread.so.0(+0x12c20) [0x7fd1a3fe1c20]
/lib64/libc.so.6(gsignal+0x10f) [0x7fd187bc137f]
/lib64/libc.so.6(abort+0x127) [0x7fd187babdb5]
/lib64/libc.so.6(+0x21c89) [0x7fd187babc89]
/lib64/libc.so.6(+0x2fa76) [0x7fd187bb9a76]
/opt/resolve/bin/resolve() [0x21718b6]
/opt/resolve/bin/resolve() [0x216fa78]
/opt/resolve/bin/resolve() [0x216f388]
/opt/resolve/bin/resolve() [0x3af00c8]
/opt/resolve/bin/resolve() [0x35a3e7d]
/opt/resolve/bin/resolve() [0x35a3d0e]
/opt/resolve/bin/resolve() [0x218e966]
/opt/resolve/bin/resolve() [0x218e8e0]
/opt/resolve/bin/resolve() [0x214658d]
/opt/resolve/bin/resolve() [0x21424bf]
/lib64/libc.so.6(__libc_start_main+0xf3) [0x7fd187bad493]
/opt/resolve/bin/resolve() [0x214153b]
Signal Number = 6

================================
Aborted


Please let me know what I can do.

Thanks.

[Dwaine Maggart]

From the CentOS Applications list, run the DaVinci Resolve Capture Logs app, and send the .tgz file it creates on the desktop.

After desinstalling/reinstalling DaVinci and created a new local database, that seems to launch 90% of the time.

However, since that, I'vet lost the connection to the DaVinci Mini Panel (PoE). The same PC, the same panel, the same firmware, the same network... that works fine on Windows (dual boot system).

Please find the logs in attachmnent

EDIT
Weird thing: launching Davinci from the command line "sudo ./resolve" brings back the control panel. Anyway, I have so much troubles, please check my logs Dwaine.

By the way, what command line do you suggest to kill completely DaVinci when it hangs?

[Dwaine Maggart]

I don't see anything obvious. It does seem to be startup hanging or crashing in the Fairlight area. But no indication of why.

Is this a brand new machine? Did you have CentOS 7 on it working previously?

The fact it's a 64 core Threadripper sort of implies a new system.

I don't have a 64 core Threadripper system to test CentOS 8.5 on.

The issue could be related to the CPU. Or to CentOS 8.5, or potentially anything else.

Is there a reason you used CentOS 8, instead of 7? 8 is EOL in a few days. 7 is not EOL until June 2024.

I'm wondering if CentOS 8 is all that supports the 64 core CPU?

You should not need to command line start Resolve with sudo. If you don't use sudo, does it start, but then not recognize the panel? If so, sort of sounds like a permissions issue of some sort. In most of the logs where Resolve is running, it appears that the panel has connected and should be working. Did you use the command line start most of the time?

I'll put together a CentOS 8.5 system and see how it behaves with my Intel CPU. Might show if there is some generic issue with Resolve and 8.5. And also see how the Mini panel behaves.

To kill a hung Resolve, use the Utilities - System Monitor app. Sort by Memory use. Either the resolve process or the GUI Thread process should be using the most memory. Right click the process and select Kill. (Not sure why the Resolve app process sometimes shows as resolve and sometimes as GUI Thread. Whichever it is, it should be using over 200 MB of memory and that's the one to Kill).

I don't see anything obvious. It does seem to be startup hanging or crashing in the Fairlight area. But no indication of why.

It was at the "Loading Fairlight" splash exactly.

Is this a brand new machine? Did you have CentOS 7 on it working previously?

It was built in February 2020. I never had Centos 7 before. I followed Seth Godin's tutorial with version 8.

Is there a reason you used CentOS 8, instead of 7? 8 is EOL in a few days. 7 is not EOL until June 2024.

When I installed Centos in February 2020, I didn't know Centos 8 was EOL in December 2021. I went to Centos 8 directly because of the very recent hardware.

I'm wondering if CentOS 8 is all that supports the 64 core CPU?

Yes it does. It was stable with the different CentOS 8.X before, except that Dolby Vision enabled makes freezing completely the system (need to turn off the computer). By the way, you should really try to work on a HDR project with DV enabled and see if there's no hang. The annoying stuff is that it occurs very randomly. It can be 5 min like 1 hour.

You should not need to command line start Resolve with sudo. If you don't use sudo, does it start, but then not recognize the panel?

Since I tried the sudo command line, the panel is recognized each time when launching DaVinci from the usual shortcut.

If so, sort of sounds like a permissions issue of some sort. In most of the logs where Resolve is running, it appears that the panel has connected and should be working. Did you use the command line start most of the time?

No, I only used it that time to try.

To kill a hung Resolve, use the Utilities - System Monitor app. Sort by Memory use. Either the resolve process or the GUI Thread process should be using the most memory. Right click the process and select Kill. (Not sure why the Resolve app process sometimes shows as resolve and sometimes as GUI Thread. Whichever it is, it should be using over 200 MB of memory and that's the one to Kill).

I can't find any task named "resolve". I'll try the GUI thread then.

Thank you very much.

[Dwaine Maggart]

@Lucius: You commented: "By the way, you should really try to work on a HDR project with DV enabled and see if there's no hang. The annoying stuff is that it occurs very randomly. It can be 5 min like 1 hour."

I updated my CentOS 8 drive to 8.5, then created a 38 clip project, enabled Dolby Vision 4, did an analysis, and just to make the GPU work a bit harder, set the timeline resolution to 8K DCI.

It's been playing for over 12 hours with no issue.

Perhaps you can put together a Project Export .dra and PM a link to it, so I can test exactly what you are having issues with.

@Lucius: You commented: "By the way, you should really try to work on a HDR project with DV enabled and see if there's no hang. The annoying stuff is that it occurs very randomly. It can be 5 min like 1 hour."

I updated my CentOS 8 drive to 8.5, then created a 38 clip project, enabled Dolby Vision 4, did an analysis, and just to make the GPU work a bit harder, set the timeline resolution to 8K DCI.

It's been playing for over 12 hours with no issue.

Perhaps you can put together a Project Export .dra and PM a link to it, so I can test exactly what you are having issues with.

Thanks for testing, Dwaine. I already sent you a HDR project in July that you tested with success also. Maybe it's related to my hardware configuration only, like something related to AMD Ryzen / Chipset. Since February 2020, I've never had any freeze with DaVinci except with these both HDR projects.

[Dwaine Maggart]

My bad. I said Project Export. I meant Export Project Archive, which makes a .dra and includes source media.

Hello @All,

i have the same issues while installing resolve 16 and 17 on debian 10 with nvidia quadro p1000.

After a lot of log readings i found a solution for the "EBADF" error.

1. create "~/.local/share/DaVinciResolve/configs/.config.data" and "touch" the following into it (make a backup, if it exist):

Code: Select all

IoFsNum = 1
IoFsMount_1 = /tmp
IoFsMappedMount_1 =
IoFsDirectIO_1 = 1
CustomLutNum = 0
Panels = none
JLCooperConfig = 192.168.254.101:23
DaVinciMiniEthPanelConfig =
DaVinciMiniEthPanelConfigIsManual = 1
SDI = none
SDIMonitorIndex = 0
SDILiveGradeIndex = 0
SDIAudioDelay = 0
RedRocketEnable = 2
RedGPUEnable = 1
EasyDCPDecoderEnable = 0
UseHwAccForDecodeEnable = 1
HardwareDecodeMask = 3
BrawGPUEnable = 1
EnableAutoFileRescan = 0
EnableDeckLinkAudio = 0
ShowAllMountedVolumes = 1
UseFloatGLPixelFormat = 0
System.Scripting.Mode = 0
EnableDisplayColorProfile = 0
Tag709ClipsAs709A = 0
ReleaseSDIOnFocusLoss = 0
VSTPathNum = 1
VSTPath_1 = /usr/lib/vst3/
DefaultVST3PathsAdded = 1
Memory = 4054
ResolveMemoryUsageInPercentage = 75
FusionMemoryUsageInPercentage = 73
GPUProcessingMode = OpenCL
GPUScopes = 1
Update.AutoCheck.Enabled = 1
Update.AutoCheck.OptInNewBetaPrograms = 0
RemoteRender.AutoScan.OtherDBs = 0
CrashReport.Enable = 1
CrashReport.AutoSend = 0


(u can change "IoFsMount_1=" to a location where u want. Please use a local folder (not a remote / NAS location)

2. Save it and start resolve. If the config window opens, then check "Media Storage" is "/tmp" and "Memory and GPU" is "OpenCL" for "CPU processing mode". If it so, then click "Save"

3. If resolve closed / crash then restart it again.

4. Now it should work.


JimJim
XUN

[Dwaine Maggart]

But it won't work well. You don't want to use OpenCL processing with an NVIDIA GPU. You need to use CUDA.

I always use CUDA.

[Dwaine Maggart]

My CUDA response was to a message that was subsequently deleted.

If you are utilizing a NVIDIA GPU, I highly encourage anyone using any flavor of Linux to run Resolve in a container (Docker and/or Podman):

viewtopic.php?f=21&t=153087

AFAICT, it's almost all upside -- almost no downsides.

Thanx for entire team which make Davinci Resolve for Linux. I use it with successful on Linux Debian flavour since two years (Debian 11 now). For installation I have used makeresolvedeb by Daniel Tufvesson. This work out-of-the-box, without any patches, only a few autoinstalled packages from Debian repositories. Apart from Resolve itself, I also use the BRAW Player. Play well!
Speed Editor also works fine.
Bertold

My platform:
Gigabyte Aorus Ultra X470 motherboard
Ryzen 2700X
128GB RAM
SSD's + HDD
Gigabyte Aorus Master RTX3070 NVidia card

Debian 11 with backports repository.
Result of `uname -a` returns: Linux 5.10.0-11-amd64 #1 SMP Debian 5.10.92-1 (2022-01-18)
Driver for NVidia from Debian backports repository: 640.91.03-1
Audio is replaced by pipewire system and works on 192kHz, comes from testing repository (0.3.40-1 version) and I am not keeping it up to date.

I don't know if any one will care but there are now experimental Fedora rocm openCL packages available. I haven't had any luck getting my RX580 going with them but they may work for some one with a 6000 series card. These will all be in the repo for the 36 release.

https://www.phoronix.com/forums/forum/l ... ost1310267

On a similar note MESA 22 has been forked and will be shipping soon. It will be adding image support and upping the openCL version to 3 so it may be a solution for people trying to get things going on Fedora. I am so sick of having to close every thing to reboot between RH and Fedora.

But do they have proper CL/GL interop ? (i.e. sharing objects between OpenCL and OpenGL)
v17 needs this to work.

I’m having a really strange problem in Ubuntu.

Resolve will sometimes start and sometimes it won’t get past the splash screen.


Problem wasn’t present a couple versions back but is def giving me
Trouble in 17.4.5. Are any other Ubuntu users having this issue?

Yes, me too. Same on CentOS too.

[Dwaine Maggart]

Is the splash screen going away, or it's hanging at the splash screen?

How long did you wait?

The times the issue occurs, are you clicking on the splash screen?

There's no splash on my side when it occurs.

[Dwaine Maggart]

You launch it, and nothing happens?