Linux 5.9 Brings Safeguard Following NVIDIA's Recent "GPL Condom" Incident

[ix900]

Open Source/AMD fans seem not to possess any argumentation skills.

The only argumentation skill they have is against Nvidia. AMD can do no wrong lol. In fact, AMD hasn't tried to put crap code into the kernel. No they never did because this is an Nvidia discussion.

[JustK]

[...] The RX 5000 series does not support HW accelerated video decoding yet via the open source Mesa stack. End of freaking story. AMD fans cannot even read - they can only posture.

% vainfo
libva info: VA-API version 1.8.0
libva info: Trying to open /usr/lib64/va/drivers/radeonsi_drv_video.so
libva info: Found init function __vaDriverInit_1_8
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.8 (libva 2.8.0)
vainfo: Driver version: Mesa Gallium driver 20.2.0-rc2 for AMD Radeon RX 5700 (NAVI10, DRM 3.38.0, 5.8.0, LLVM 10.0.0)
vainfo: Supported profile and entrypoints
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileVC1Simple : VAEntrypointVLD
VAProfileVC1Main : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointVLD
VAProfileH264Main : VAEntrypointEncSlice
VAProfileH264High : VAEntrypointVLD
VAProfileH264High : VAEntrypointEncSlice
VAProfileHEVCMain : VAEntrypointVLD
VAProfileHEVCMain : VAEntrypointEncSlice
VAProfileHEVCMain10 : VAEntrypointVLD
VAProfileHEVCMain10 : VAEntrypointEncSlice
VAProfileJPEGBaseline : VAEntrypointVLD
VAProfileVP9Profile0 : VAEntrypointVLD
VAProfileVP9Profile2 : VAEntrypointVLD
VAProfileNone : VAEntrypointVideoProcvainfo

[eydee]

This only hurts the users. Both sides should be ashamed. Fuck politics.

[skeevy420]

Also the part where there is a proprietary AMDGPU-PRO driver that is not FOSS but tends to be a bit ahead in terms of compatibility and stability. Not sure why birdie would bother with changing hardware, seems like poor troubleshooting on his part. My Vega56 card is rock solid stable on Ubuntu 20:04 for Steam gaming, and it does all the things you would expect, like video decoding and encoding, etc.

That doesn't help people of non-Pro distributions. I don't know if he mentioned the distribution used.

How long have you had your Vega56 for? IIRC, they didn't have the best reputation around these parts for the first year or so. Just curious if you had to go through, for a lack of a better term, the growing pains of the driver being developed or if you bought in after a year of so. Anecdotally, it seems that people who go through the growing pains are more critical of AMD than either people who bought in afterwards and didn't deal with the major bugs or the general AMD Linux enthusiast who wants them to succeed.

I've had my Polaris card since February last year. An RX 580. Would have bought it a few months sooner but that was when GPU inflation hit and I refused to pay those prices unless forced. By the time I bought it all the major platform bugs and kinks had been worked out during the 4XX Polaris generation so I was able to buy into a refined platform with a more mature driver where my biggest issue is software would say RX 480 instead of RX 580. I'd have held out a few months longer for GPU prices to go back down but my 260x was 7 years old and started getting coil whine with random crashes so I was forced into an upgrade and I'm damn glad it happened. I'm in the category of enthusiast who wants them to succeed but also reads tech blogs and doesn't buy day one hardware.

Oh snap, just realized I can finally start downloading the THPS Warehouse demo

[Volta]

So people pushing ideologies are once again making life harder for people who just want to use their hardware.
I'm surprised BSD isn't getting more traction due to this, to then eventually just give up on Linux. Would save a lot of headaches.

If someone is making others life harder in this case it's nvidia. It's funny you want more permissive license than GPL, but you have nothing against proprietary blob same time. AMD, Intel already saves us from headache. Nvidia can go to hell. From what cave such hypocrites came from?

[Spacefish]

I have a 5700XT and run Kernel 5.8, the "amdgpu" driver is quite stable for me. Performance is good, no crashes..
I had random blackscreens with older kernel though!
Only problem that persists: Monitor stays black when i turn of the monitor and turn it on again.. However sleep / standby / wakeup works!
Although i had similar problems with a GTX 970 on windows with the same monitor, so the monitor not behaving correctly regarding display port protocols could be a factor here.

[schmidtbag]

In the same way that Wayland failed relative to X-Windows ?

I guess so. I personally don't think Wayland is a failure, but I also don't think BSD or Haiku are either. ReactOS is kinda a failure though lol.

[andre30correia]

I have a 5700XT and run Kernel 5.8, the "amdgpu" driver is quite stable for me. Performance is good, no crashes..
I had random blackscreens with older kernel though!
Only problem that persists: Monitor stays black when i turn of the monitor and turn it on again.. However sleep / standby / wakeup works!
Although i had similar problems with a GTX 970 on windows with the same monitor, so the monitor not behaving correctly regarding display port protocols could be a factor here.

one after the card arrive right? Amd drivers are really bad, since old days, for some reason people who wants stable drivers buy nvidia

[unic0rn]

https://bugzilla.kernel.org/show_bug.cgi?id=207441

https://bugzilla.kernel.org/show_bug.cgi?id=207735

https://bugzilla.kernel.org/show_bug.cgi?id=207881

https://bugzilla.kernel.org/show_bug.cgi?id=207637

3 related to fans (possibly depends on exact card model and firmware), 1 caused by using old kernel.

i mean, what kind of sane person runs navi with 5.6? don't get me wrong, i have nothing against nvidia, i wish them well. but switching from amd to nvidia because of problems caused by outdated kernel and/or mesa, then throwing sh.. at amd on the forum is just silly at best, trolling at worst.

Not a single person on Earth is able of auditing the entire source code of the Linux kernel and proving it does not have backdoors or malicious functions. Again, sanity is not strong with this one. You're absolutely delusional in pretty much everything you're saying.

your attitude shows. code is inspected and tested before being released. what's in the kernel, was audited. if you don't like and/or trust the linux kernel team, you may as well switch to another operating system instead of wasting your time calling people delusional when it's your behaviour that rises some questions.

Buying an AMD CPU along with an AMD GPU and finding critical bugs (shown in the thread earlier) in the AMD open source drivers is shilling for NVIDIA. Right. Open Source fans have no common sense and logic.

bugs that were possibly already fixed but you just didn't bother to use a recent enough kernel and/or mesa? where is your common sense?

% vainfo
libva info: VA-API version 1.8.0
libva info: Trying to open /usr/lib64/va/drivers/radeonsi_drv_video.so
libva info: Found init function __vaDriverInit_1_8
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.8 (libva 2.8.0)
vainfo: Driver version: Mesa Gallium driver 20.2.0-rc2 for AMD Radeon RX 5700 (NAVI10, DRM 3.38.0, 5.8.0, LLVM 10.0.0)
vainfo: Supported profile and entrypoints
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileVC1Simple : VAEntrypointVLD
VAProfileVC1Main : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointVLD
VAProfileH264Main : VAEntrypointEncSlice
VAProfileH264High : VAEntrypointVLD
VAProfileH264High : VAEntrypointEncSlice
VAProfileHEVCMain : VAEntrypointVLD
VAProfileHEVCMain : VAEntrypointEncSlice
VAProfileHEVCMain10 : VAEntrypointVLD
VAProfileHEVCMain10 : VAEntrypointEncSlice
VAProfileJPEGBaseline : VAEntrypointVLD
VAProfileVP9Profile0 : VAEntrypointVLD
VAProfileVP9Profile2 : VAEntrypointVLD
VAProfileNone : VAEntrypointVideoProcvainfo

shhh, the fact that open source software makes progress may blow someone's mind.

now, i'm not saying there are no bugs - far from it. heck, i'm running kaveri and i had some problems months ago, seemed to calm down after switching from 4.19 to 5.6 and some mesa updates - and that's a pretty old hardware already. but to be fair, those were mostly caused by me testing every possible option to reach maximum performance, like gl_threads, or caused by the combination of dxvk and a game screwing up resource management, like gta v running out of vram (i got rid of hangs caused by gta v by changing amdgpu.gttsize and limiting vram usage with dxvk.conf - that just shows how many factors are there to each possible crash).

with opensource drivers, running the latest kernel and mesa with default settings is the first step to eliminate problems. also, at least on kaveri i've found radv to be more stable than amdvlk - the latter was causing some non-critical kernel errors in some games, whereas radv is just stable and works as supposed. so while i 100% agree that it's not all bug-free, the actual quality of the drivers is pretty high.

and disclaimer, if i were to chose a dedicated gpu right now, i would pick nvidia, simply because at the <=75W range amd has nothing reasonable. hopefully navi 2 will change that, and hopefully nvidia will change their attitude towards linux kernel. bottom line is, each choice has its drawbacks and each driver has its problems. "shilling flamewars" are not a solution.

[NotMine999]

I noticed that almost all of this entire thread went sideways into graphics card support and gaming simply due to the mention of NVIDIA. I guess that happened due to the original article requiring deeper diving to understand it's actual topic. So this article is "unintentional clickbait" IMHO.

I wonder if anyone bothered to look over the actual code that was submitted to see the use case being considered? I did. It looks like a machine learning application that uses GPUs for the processing and Mellanox for the networking. It's an approach to offload networking processing from the CPU to the GPU in this use case, which seems reasonable to consider.

Could it be leveraged into other networking things? Perhaps.

Is this code evil? You'll have to define your subjective feelings on that in an objective manner to intelligently argue them. Just because "IT'S NVIDIA!!" is not enough.

Do I agree with the comments by Greg KH about the quality of the submitted code? Yes, it is SLOPPY by kernel standards (when you have to tell a submitter to run 'checkpatch' before submission /facepalm/) and in other ways (the entire GPL symbol exporting code in it is a hot mess), but it doesn't mean this technical area (high bandwidth, HW accelerated networking for machine learning) should not be investigated for inclusion in the Linux kernel.