Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
Graphics DRM Linux

Linux 5.19 Adds 500K Lines of New Graphics Driver Code (phoronix.com) 79

Posted by BeauHD from the new-and-improved dept.
UnknowingFool writes: The current Linux kernel in development, 5.19, added 495,793 new lines of code for graphic driver updates. David Airlie sent in the new lines as part of Direct Rendering Manager (DRM) subsystem of Linux. The majority of additions were for AMD's RDNA and CDNA platforms but Intel also submitted changes for their DG2 graphics as well. Updates also came from Qualcomm and MediaTek for their GPU offerings.
This discussion has been archived. No new comments can be posted.

Linux 5.19 Adds 500K Lines of New Graphics Driver Code More

Linux 5.19 Adds 500K Lines of New Graphics Driver Code

Comments Filter:

  • I don't see the problem (Score:4, Informative)

    by gweihir ( 88907 ) on Thursday May 26, 2022 @07:32PM (#62569256)

    Sure, adding lots of code is always something that needs a good reason and needs to be done carefully or the project will go to hell. That said, all this code is completely optional on a headless server and will not even be present if basic hardening is done. And on a reasonably hardened desktop you will only compile or install the modules for the hardware actually present. So not a problem at all.

    • Yeah, plus the kernels very modular. If its presenting a problem, just switch it out for a different module. Or in the case of headless, dont even load that module at all.

      • Re: (Score:2)

        by gweihir ( 88907 )

        Indeed. Good design overall.

      • Re:I don't see the problem (Score:5, Informative)

        by Antique Geekmeister ( 740220 ) on Friday May 27, 2022 @02:17AM (#62569776)

        Few people produce minimized kernels anymore. It's rarely worth the extra testing time and support time, especially when some bit of software is unexpectedly reliant on a specific graphics driver. Being headless does not mean that a server may not transform graphics images, for example.

        • True. I can't remember the last time I recompiled a kernel. Nowadays, it's rare to even modprode anything.

        • Re: (Score:2)

          by gweihir ( 88907 )

          Few people produce minimized kernels anymore.

          Well, yes. Many people are indeed completely incompetent when it comes to IT security. There are reasons why the current ransomware epidemic is so massive and starts to affect Linux as well. Incidentally, in regulated environments, you are often _required_ to harden some systems and that means minimizing the kernel among other things.

          • Spending developer or sys-admin time optimizing kernels is an excellent example of what Donald Knuth referred to as "premature optimization". If I may quote him, “The real problem is that programmers have spent far too much time worrying about efficiency in the wrong places and at the wrong times; premature optimization is the root of all evil (or at least most of it) in programming.”

            Recompiling kernels is fragile. It's very easy to fracture a stable system with the wrong compilation environme

    • On a desktop system, I don't want to mess around with anything. I just want to use it as it is. I've been working exclusively with linux on servers and desktops since 1998, and at this point in life I don't want to tinker with anything that I don't need. In early 2000s you had to tinker with your desktop to make it work, servers were ok, and that was fun. Now it's just a drag.

      Anyway, they did not just add code... they added half a million lines of code (not improvements or features, but DRM) and that's con

      • Re:I don't see the problem (Score:4, Informative)

        by serafean ( 4896143 ) on Friday May 27, 2022 @04:14AM (#62569848)

        > not improvements or features, but DRM

        Um, you do know that in this case DRM means Direct Rendering Manager, ie the kernel the kernel's GPU subsystem (written in the article). So in effect all this is new features (supporting new hardware). And a lot of the code is autogenerated headers for various hardware revisions.

        • Um, you do know that in this case DRM means Direct Rendering Manager, ie the kernel the kernel's GPU subsystem (written in the article).

          Indeed, very funny to have picked up that name and acronym from the Linux guys.

        • Not only that but the graphics updates were for many different systems like desktop for AMD, laptops for Intel, phones for Qualcomm, and embedded devices for Mediatek SOCs so it is unlikely that the additions appear in the same hardware.

        • Re: (Score:2)

          by sconeu ( 64226 )

          Not only is the acronym explained in TFA, it's explained in TFS!!!!

        • Re: (Score:2)

          by gweihir ( 88907 )

          > not improvements or features, but DRM

          Um, you do know that in this case DRM means Direct Rendering Manager, ie the kernel the kernel's GPU subsystem (written in the article). So in effect all this is new features (supporting new hardware). And a lot of the code is autogenerated headers for various hardware revisions.

          Hehehehe, indeed. Spotted the clueless one.

      • Re:I don't see the problem (Score:5, Funny)

        by DamnOregonian ( 963763 ) on Friday May 27, 2022 @04:59AM (#62569874)
        You use Linux as your main OS, and you didn't know what DRM meant in the context of graphics drivers?

        Shit, if that's not proof that it's the year of the Linux Desktop, I don't know what the hell is.

        • Re: (Score:2)

          by gweihir ( 88907 )

          You use Linux as your main OS, and you didn't know what DRM meant in the context of graphics drivers?
          Shit, if that's not proof that it's the year of the Linux Desktop, I don't know what the hell is.

          Sad as it is, I agree that the massive influx of the clueless into the Linux-space is a good indicator.

          • I'll be damned. People are using Linux desktops without knowing the bloody fucking internals of the graphical system, and you and I agreed on something... The world is looking up.
  • How hard does it gotta be to integrate NVIDIA drivers into Linux, for God's sake?!

    • Re: (Score:3, Informative)

      by NateFromMich ( 6359610 )

      How hard does it gotta be to integrate NVIDIA drivers into Linux, for God's sake?!

      Are you suggesting that everyone should work extra hard, entirely on their own, to get some closed source crap working because you like Nvidia?
      They recently announced that they are finally going to contribute, so let's see how that plays out.
      Bitching that a bunch of people working on their own aren't furthering the desires of an open source hostile company like Nvidia doesn't seem useful.

      • Are you suggesting that everyone should work extra hard, entirely on their own, to get some closed source crap working because you like Nvidia?

        More likely because the support provided by the open source nouveau driver is so bad for so many Nvidia devices as to be practically useless, and also the fact that many systems out there ship with Nvidia hardware.

    • I believe they recently announced work on an open source version of their driver.

      • Also, right now there is an open source accelerated driver in Linux for nvidia called "nouveau".

        • And it's fantastically terrible unless you're working with a desktop with a 3+ year old GPU in it. Everyone else installs proprietary drivers from Nvidia, or has to blacklist nouveau just to get their Optimus-equipped laptop to boot up at all, or both.

    • NVidia keeps modifying the suite of kernel and of userland graphical libraries that interact with the kernel. So it's not as easy as you might think. It's especially difficult because they'd need to publish it as GPL, which means the secret tuning to cheat in well-known graphics benchmarks would be apparent to all.

      • "secret tuning" to "cheat benchmarks" TOP FUCKING KEK. With the droves of tech-based review sites/youtube channels no company would get away with fudging benchmark numbers without immediately getting called out for chuckle-fuckery. A quick Bing search yielded 0 results for GPU benchmark cheating, but tons of iPhone and Android devs doing it. The last GPU I purchased (RTX2080ti) actually SURPASSED the documented performance numbers out of the box.

        But this has nothing to do with why Nvidia drivers for linux

        • NVidia has repeatedly been caught cheating at benchmark tests. They were caught cheating in the TPCx-BB tests last year

          * https://www.theregister.com/20... [theregister.com]

          I'm confused why you think "they wouldn't do it because they'd be caught" reliably prevents abuse, or even criminal fraud, involving benchmarks. That's specially a risk if there is a long delay between cheating the benchmark and getting caught at it. The benefits are very high and the penalty for getting caught are often accepted as part of the cost of doi

      • Publish it like this [nvidia.com]?

        Oh hey lookie there - GPL and MIT licensing for open-source Nvidia kernel modules, announced and released a week ago. Feel free to dig through the source to find proof of your accusations [github.com] or shut the fuck up.

        I just love it when people bitch about companies not doing "the right thing" when they actually finally have.

        • The rest of the necessary tools don't necessarily have such licenses. A check of the libEGL libraries shows better licensing, but no updates in 4 years. Assembling all the necessary bits to provide the full suite is not not as certain or as thorough as we might want in open source development, and some skepticism is justified.

    • Define driver. Do you want 3D graphics? Seems to work just fine here. Do you want up to date weekly patched game ready drivers for high performance gaming? Then you'll need to convince more people to actually game on Linux before a vendor will deem that worthwhile.

      Or do you just want everything open source due to some sense of purity or entitlement over someone else's commercial offering?

      • This.

        Plus AMD already has open source drivers, if that's so much better of a route to go then support them. I don't understand how people are so keen to support companies that don't support them and then whine about it while shunning the companies that do support them. If AMD is already giving you everything you're asking for why not show them some support? Or is it some sense of necessity reach out and help poor Nvidia from hurting themselves by making bad decisions?

        • Nvidia disagrees that it's a "bad" decision, since they decided to open source their driver [github.com] a week ago.

          What else have you got?

          • Nvidia disagrees that it's a "bad" decision, since they decided to open source their driver [github.com] a week ago.

            I didn't say open sourceing the kernel mode component was a bad decision, clearly they don't think it was a bad decision otherwise they wouldn't have done it. Of course the user-land component is closed (as that includes their proprietary technologies like the CUDA implementation) and the open component is only for Turing and Ampere architectures. All I'm saying is AMD already has a completely open driver stack so if you really want that then it already exists (and has done for some time).

    • How hard does it gotta be to integrate NVIDIA drivers into Linux, for God's sake?!

      Linux + NVIDIA is pretty much the only way to go for GPU-based HPC in the datacenter, it works extremely well. On the desktop/client side the loadable kernel module can be automatically recompiled via DKMS when the kernel is updated and the ABI changes.

    • My guess? The first problem is that programming for Linux is a moving target, to this day it does not have a stable API and the functionality you count on to make your driver work may not exist next month.

      And the second problem is that you cannot open your driver to the general public if you have proprietary code in it, or code covered by commercial agreements between your company and other companies and where almost certainly your company has paid a lot of money to be able to use this code. I won't even

      • I know this is Slashdot, and nobody bothers with reading anything or keeping up on current status of things before spouting off, but you do know that Nvidia open-sourced their Linux driver last week [github.com], right?

        The fact that repo exists, as well as AMD and Intel GPU driver repos, and they are public basically shoots your argument in the testicles.

        • You must be a famous party guy with your orangutan upbringing, right?

          And I suspect that you also don't know how to interpret what you read, since you didn't realize that the driver that nVidia made available, although in fact more "open" than the previous ones, still depends on a closed binary provided by nVidia to work. For me personally it's not a problem, but the free source folks consider that this isn't really open source yet.
    • Never ask that

    • Well considering that Nvidia just open-sourced the core of their drivers [nvidia.com], it's a lot less hard than it used to be. Or, at least, will have much tighter integration in the next kernel version as it was probably too late for this one.

  • Nobody's freaking out about the term "DRM" this time! Just goes to show what happens when you carefully define the initialism in the summary...

    • Re: (Score:3)

      by Entrope ( 68843 )

      You must have screened put this AC comment [slashdot.org] that complained about the name.

      • Or this one:
        https://tech.slashdot.org/comm... [slashdot.org]

      • Re: (Score:2)

        by Opyros ( 1153335 )
        Well, that wasn't quite what I meant about "freaking out". It used to be that whenever the DRM subsystem was mentioned here, there'd be people who flew into a panic about Digital Rights Management being added to Linux. But I see I spoke too soon; as samwiche points out, someone did make that very error several hours after my other comment! Uh, thanks, Slashdot, for not disappointing me?

    • Nobody's freaking out about the term "DRM" this time!

      Why would they? The term dates back to before the millennium bug.

  • I never saw the problem with closed sourced drivers loaded from blobs by some shim. Any functionality I get from a free OS I consider a blessing. Then I boot to Windows because an apt update blackscreened the x manager again and I just need to get stuff done :|

  • Shouldn't that stuff be orthogonal to the kernel?

Slashdot Top Deals

Receiving a million dollars tax free will make you feel better than being flat broke and having a stomach ache. -- Dolph Sharp, "I'm O.K., You're Not So Hot"

Close