Announcement

Collapse
No announcement yet.

KDE On Wayland: "The Biggest Thing Needed Now Is Adoption By 3rd Party Apps"

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Originally posted by mSparks View Post
    Nope, Xorg server is completely independent of logins, applications or any work you are doing, unlike the wayland compositors it doesn't even need to run on the machine you are working on.
    Its not a wrong assume. He is writing up how it behaves on your normal Linux Desktop setup. X11 applications normally don't run if there is no Xserver to connect to.

    Originally posted by mSparks View Post
    Probably he's experiencing AMD GPU driver crashes, they will take out most of everything else, and no amount of "robustness protocol" is going to recover from a CPU not executing any instructions, production Nvidia and Intel drivers basically never crash.
    The reality is both Nvidia and Intel graphics drivers crash.

    These AMD GPU crashes. Guess what KDE with robustness manages to keep on going. Not ideal to be in a driver crash loop. Yes those using gnome and x11 were finding themselves back at login screen again not hard locked. Power button triggering a correct shutdown again not hard locked. Also note CRIU support. Final part of robustness protocol is the means to suspend applications to disc.

    AMD and Intel with problem normally don't hard lockup. If the problem is really bad you end up in loops of driver crashing and restarting with AMD and Intel.

    Please note DRI2 and before AMD/Intel drivers also use to hard lockup. Its the DRI3 drivers that introduced per process memory that saw hard locking stop.

    AMD CPU not running instructions any more
    After I ssh into my Ubuntu 20.04 machine, I quickly start getting the error message kernel:[ 632.398797] watchdog: BUG: soft lockup - CPU#17 stuck for 22s! [irq/119-nvidia:1148] in my terminal every 10 seconds or so. Nothing I tried seemed to fix it, so I actually did an entire fresh installation of Ubuntu. As soon as I installed the Nvidia driver, I had the issue again (on the same CPU, #17) Using top, I can see that irq/119-nvidia is using 100% of a cpu core. Any ideas what this could be? ...

    That a Nvidia driver feature caused by the driver hard locking up.

    Originally posted by mSparks View Post
    As clearly evidenced by you having to go all the way back to 2017 to find someone having a desktop related issue?
    Funny. If not a little sad.
    We are running the following setup Ubuntu 22.04 - 5.19.0-41-generic Dual A100 Cards PCIE NVIDIA driver - 525.85.07 KVM/Libvirt OpenStack ZED We have both GPU set to MIG mode, with the following slices 1x 3g.40Gb 1x 2g.20Gb 4x 1g.10Gb Everything works as expected apart from now and again when a user destroys/creates a vgpu backed instance the following occurs, usually resulting in a hardlockup of the entire hypervisor. CPU and load avg increases significantly DMESG reports stuck CPU,...

    No I did not want to having you under so many links that is not funny so I picked one of the oldest. Yes I have Nvidia driver caused hard lockups from 2017 to 2023 with multi cases.

    Originally posted by mSparks View Post
    ​That AMD bug is pretty much universal and current, the bug you pulled was one person and definitely resolved 5 years ago.
    Did nvidia remove the feature from their driver to that causes it to hard lockup the system 5 years ago the answer is no they did not. Nvidia only fixed the trigger event not the feature that makes it worse.

    AMD driver even the current is not hard locking. Not hard locking means applications like firefox and so on can be terminated correctly and save their state. Libreoffice and so on can save their state. The system can in fact power down while the AMD or Intel gpu is stuck in crashing loop this also avoids file system issues.

    Note the 2023 effects those using Nvidia in data centers.

    There are a lot of Nvidia driver crashes Nvidia is not magically immune to them.

    1. Issue or feature description Running "sudo systemctl start nvidia-docker causes the machine to freeze. I cannot even ssh in from another machine. 2. Steps to reproduce the issue As above. I'm so...


    Yes the Nvidia driver hard lockup problem is not only restricted to the people playing games.

    Originally posted by mSparks View Post
    old machines people aren't otherwise using any more that don't support or work well on windows, do support Linux, there is not suddenly a rush of people buying ~10 year old RX480 GPUs
    There is a problem with that theory . RX480 appeared in steam numbers with lots of users in 2016​ when the card was new with Linux. Really were do you find a 10 year old RX480 thinking it was only made in 2016.

    The change is 10 years old being the DRI3/DMABUF of 2013 but then there is a few years for that to make it out into distributions and users to notice stability difference that why the uptick is in AMD GPU sales for Linux users in 2016. Having the driver be able to restart instead of hard lock the system makes quite a difference.

    There was a change in Linux user GPU buying behavior at the time of the RX480 being new. The change clearly shows users prefer to have graphics drivers that don't hard lockup.

    Yes the rage about Nvidia drivers being broken starts around the same time when users notice the difference. Yes when nvidia was competing against DRI2 drivers having hard lockups was status normal. Nvidia needs to lift there driver crash handling to be equal to AMD and Intel GPU crash handling.

    Comment


    • Originally posted by oiaohm View Post

      Its not a wrong assume. He is writing up how it behaves on your normal Linux Desktop setup. X11 applications normally don't run if there is no Xserver to connect to.
      your computer can't start if there is no boot drive to connect to either, what relevance and value is this knowledge?
      You wont be able to start and use Wayland applications without a wayland server either.
      Originally posted by oiaohm View Post
      The reality is both Nvidia and Intel graphics drivers crash.

      These AMD GPU crashes. Guess what KDE with robustness manages to keep on going.
      And that has been the case for the last like 30 years wayland or no wayland? so what?

      Originally posted by oiaohm View Post

      There is a problem with that theory . RX480 appeared in steam numbers with lots of users in 2016​ when the card was new with Linux. Really were do you find a 10 year old RX480 thinking it was only made in 2016.
      The RX480 is the second most popular "linux gamer" card on linux, the first being the steam deck


      and click "video card description".

      Excluding the steam deck, its 19% AMD to 16% NV to 10% Intel, which says a lot considering that market is mostly "throw away" PCs.

      Originally posted by oiaohm View Post
      Yes the rage about Nvidia drivers being broken starts around the same time when users notice the difference. Yes when nvidia was competing against DRI2 drivers having hard lockups was status normal. Nvidia needs to lift there driver crash handling to be equal to AMD and Intel GPU crash handling.
      Really, this rage is all in your head, presumably emanating from people mad at nvidia for not doing work of no commercial value for free.
      Maybe if Redhat/IBM did less work of no commercial value for free they would maybe stop hemeraging cash and see growth in users, revenue and adoption of their technologies (like, e.g. Oracle and SUSE).

      Adobe recently launched:


      Running 100% on linux nvidia gpus.
      I suspect they are less rage and more relax on luxury yachts.

      If you expect the market to adopt wayland, a compositer with little to no commercial value, that after a decade and a half still isn't even close to as good as what it was supposed to replace, with the entire dev team still only tests it on GPUs no one uses.
      You are going to have to do a lot better than an alphabet soup of unrelated technologies no one cares about.
      Maybe then they could kit you out with some DGX100s.
      How did you get on with your unix philosphy and everything is a file homework, have you at least managed to acquire a basic understanding of Linux yet?
      Last edited by mSparks; 03 October 2023, 08:20 PM.

      Comment


      • Originally posted by mSparks View Post
        You wont be able to start and use Wayland applications without a wayland server either.
        This is wrong. Wayland application with robustness can in fact start without wayland server. All wayland applications with robustness need to start is a wayland socket that is held. The difference between a direct KMS/DRI application vs Wayland application is almost no difference. Yes a Wayland application that has robustness that has started because Wayland socket is held can in fact be using the GPU rendering stuff without Wayland compositor/server started.

        Wayland robustness changing things massively. Yes you can with robustness start up Wayland compositor when you want to see the output and shut the compositor down when you don't want to see the output. As long as the wayland socket remains the Wayland application believes it to keep on running. This comes very different to X11 where the xserver of some from has to be their somewhere for X11 application to run.

        Only reason why you have not been able to start Wayland applications without wayland server of some form is lack of robustness code. You can in fact have systemd hold the wayland socket and that be enough for wayland applications with robustness to start.

        mSparks wayland robustness means you don't need the overhead of X11 server or Wayland compositor when you don't need to see the output. Held socket is basically the ultimate black-hole for output.

        Originally posted by mSparks View Post
        And that has been the case for the last like 30 years wayland or no wayland? so what?
        No it not. AMD and Intel drivers before DRI3/DMABUF on Linux don't have a restart function. So things have changed in the last 10 years and Nvidia has not caught up.

        Originally posted by mSparks View Post
        Excluding the steam deck, its 19% AMD to 16% NV to 10% Intel, which says a lot considering that market is mostly "throw away" PCs.
        There is a problem here go and compare that to Windows machines in the steam-survey where those machines have to be coming from if they are throw away PCs. Dropping from 75% to 35%. Also you missed AMD and Intel that are both mesa so combined are sitting on 64%. The market presence is on the path to being fully inverted between Windows and Linux. This divergence starts 2013.

        Yes you do have to exclude the steamdeck because its really a console where the user if they want it don't have any major choices.

        Originally posted by mSparks View Post
        Really, this rage is all in your head, presumably emanating from people mad at nvidia for not doing work of no commercial value for free.
        Not quite Redhat and other distributions have got sick of maintaining large sections of X11 server with exactly 1 user being Nvidia. Linux Distributions are sick of taking care of code Nvidia is abstraction commercial value from and having to deal with the cost of fixing security issues in it all the time.

        Originally posted by mSparks View Post
        Adobe recently launched:.
        Really.
        Today, Adobe and NVIDIA, longstanding R&D partners, announced a new partnership to unlock the power of generative AI to further advance creative workflows.

        Nvidia coded so of course it using Nvidia hardware. That is not a sign that the Linux world need to bend knee. Of course those pushing Nvidia will reference a lot work Nvidia has directly funded then nicely skip over that Nvidia using deprecated code in the Linux kernel and X11 x.org server that they are not willing to pay to have maintained. Yes every other party has updated their code not to need it.

        Linux distributions Oracle and SUSE are also on the path to Wayland and also wish to end X11 x.org server as you know it because they cannot keep on justifying the maintenance bill.. They are sick of the same problem.
        Last edited by oiaohm; 04 October 2023, 06:56 PM.

        Comment


        • Originally posted by oiaohm View Post
          This is wrong. Wayland application with robustness can in fact start without wayland server. All wayland applications with robustness need to start is a wayland socket that is held. The difference between a direct KMS/DRI application vs Wayland application is almost no difference. Yes a Wayland application that has robustness that has started because Wayland socket is held can in fact be using the GPU rendering stuff without Wayland compositor/server started.
          It is not wrong. I even demonstrated it to you with example code a few days ago.

          Originally posted by oiaohm View Post

          There is a problem here go and compare that to Windows machines in the steam-survey where those machines have to be coming from if they are throw away PCs. Dropping from 75% to 35%. Also you missed AMD and Intel that are both mesa so combined are sitting on 64%. The market presence is on the path to being fully inverted between Windows and Linux. This divergence starts 2013.

          So you think the RX480 being useless on windows in 2023 means that owners of them will start developing wayland apps for KDE?

          that's quite a leap of faith.

          And a very strange target market.
          Last edited by mSparks; 05 October 2023, 12:24 AM.

          Comment


          • Originally posted by mSparks View Post
            It is not wrong. I even demonstrated it to you with example code a few days ago.
            Your example application did not have robustness code and you had not attempted to have the bare min in existence..

            Remember xwayland has robustness code that were I found you could just create a socket that basically /dev/null and tell xwayland that this is the wayland socket and have xwayland work so that X11 applications work and you are able to screen capture from xwayland using X11 methods and have opengl/vulkan acceleration for X11 applications. Yes while totally not connected to any input or output.

            mSparks what is the barest min wayland compositor you can create that enough for Wayland applications with robustness to fire up.

            You are creating barest min wayland compositor its in fact less than 100 lines of code.
            1) setup unix socket equal to /dev/null
            2) export the environmental vars telling the application where the Wayland socket is being this null socket.
            3) go into wait for ever mode.

            Yes the barest min wayland compositor gives hello world example run for it money. Of course you can create this using systemd socket activation support and many other items like it that are not compositors as well.

            Fun feature of Wayland protocol the compositor is not required with robustness code ever to answer any of the Wayland applications requests. Robustness code the socket is there its not requesting anything the compositor must be starting up. Application is free to start using the opengl/vulkan... while the compositor is starting up like egldirect.

            mSparks create a X11 server than can provide enough that X11 application can run in under 100 lines of code.

            The concept of the bugger all headless wayland compositor goes back to early days of Wayland.

            Remember mSparks how wayland is based around EGL.


            The EGL API, along with the EGL_EXT_platform_device extension, makes it possible to create off-screen buffers and OpenGL rendering contexts on a GPU by using a DRI device rather than a 3D X server. As of version 3.0, VirtualGL supports this method of GPU access. Since the EGL API does not support all of the features of the GLX API, it is necessary to emulate a subset of the GLX API rather than simply modifying GLX commands and marshaling them to the 3D X server. Thus, whereas the server can have fewer moving parts when VirtualGL uses the EGL API, the GLX interposer is more complex and more likely to encounter application compatibility issues.​
            Wayland applications are in fact free to start creating buffers and doing rendering by EGL/Vulkan method if they are robust while they are waiting for the Wayland Compositor to start. Of course Wayland robustness means Wayland application can in fact wait forever for the Wayland compositor to functionally start up. This means all you need to-do is just the bare min that the robust wayland application believes in the possibility of a Wayland compositor appearing not that you ever make it appear.

            This is why you are wrong here mSparks. Wayland robustness code is a big deal really because things don't function as you would historically expect. Wayland robustness does give the means to start Wayland application before you start Wayland compositor as well.

            Comment




            • Originally posted by oiaohm View Post
              mSparks what is the barest min wayland compositor you can create that enough for Wayland applications with robustness to fire up.
              Code:
              /*
               *  This is a sample X11 program using Xt and the Athena widget set that
               *  simply puts up a main window with the text "hello" in it.
               */
              
              #include <X11/Intrinsic.h>
              #include <X11/StringDefs.h>
              #include <X11/Xaw/Label.h>
              
              int main(int argc, char* argv[])
              {
                  XtAppContext app_context;
                  Widget toplevel, hello;
              
                  toplevel = XtVaAppInitialize(
                      &app_context,
                      "XHello",
                      NULL, 0,
                      &argc, argv,
                      NULL,
                      NULL);
              
                  hello = XtVaCreateManagedWidget("hello", labelWidgetClass, toplevel, NULL);
              
                  XtRealizeWidget(toplevel);
                  XtAppMainLoop(app_context);
              }​
              Then the user specify an Xwayland display as the display, most of the robustness delivered by the TCP/IP stack

              Dunno, why you would want to though, the performance is horseshit. we benched it 5 months ago

              Comment


              • Originally posted by mSparks View Post
                Attempted Move of goal post deleted.
                We have had this debate before you lost it then and I am not starting this one. You just brought up demo of Nvidia bug nothing more.

                Comment


                • Originally posted by oiaohm View Post

                  We have had this debate before you lost it then and I am not starting this one. You just brought up demo of Nvidia bug nothing more.
                  there is no bug there afaict, wayland is "working" just fine.
                  it is just not performing as well as X11, in any respect, on any gpu, for any application.

                  And if it is a nvidia bug:
                  Personally I don't share your confidence that nvidia will get round to fixing bugs in the wayland source code before the 12 month deadline wayland now has for 3rd parties to adopt it.

                  I will of course revisit it for the next large core project in 5 or 6 years. Maybe the situation will have improved by then.

                  Comment


                  • Originally posted by mSparks View Post
                    And if it is a nvidia bug:
                    There is a Nvidia developer allocated to the Xwayland under performance with nvidia drivers I did give you the Nvidia forum link for issues where the Nvidia developer stated he working on.

                    This is absolutely infuriating. Even more infuriating is people on all support/community avenues (reddit, phoronix, gits, here) blaming everything else but the nvidia proprietary linux driver. On x11 the driver literally can’t do anything in parallel without shitting itself to stutters, severe frame drops, judder, and even audio hitches. Just hovering the mouse over buttons, or a popup showing up, or opening an app, or a video playing on browser or mpv and you’re guaranteed to drop frames on e...

                    Remember when you said Wayland had performance issues. Guess what Nvidia has performance issues under X11 as well when KMS mode is used by Nvidia drivers under X11.

                    The Nvidia closed source driver has issues quite major one. Wayland compositor just happens to force KMS/DMABUF usage on. Yes Nvidia X11 driver can at times automatically turn that feature on and let loss hell.

                    The performance issues with Wayland are purely inside Nvidia driver code you can replicate them under X11 using Nvidia closed source driver by forcing particular options on or if you are unlucky enough to have the driver turn those options on by default or because you have used gamescope or other wayland compositor under X11 and nvidia drivers decide to switch over to the KMS code..

                    Originally posted by mSparks View Post
                    it is just not performing as well as X11, in any respect, on any gpu, for any application.
                    That the bug degraded performance. In fact the nvidia closed source driver with X11 with right options in fact slows down even more than running wayland compositor

                    Comment


                    • Originally posted by oiaohm View Post

                      There is a Nvidia developer allocated to the Xwayland under performance with nvidia drivers I did give you the Nvidia forum link for issues where the Nvidia developer stated he working on.
                      All the unresolved issues nvidia lists for wayland and xwayland are marked as protocol design flaws or compositor bugs.
                      You should know, you liked it.
                      Originally posted by oiaohm View Post

                      This is absolutely infuriating. Even more infuriating is people on all support/community avenues (reddit, phoronix, gits, here) blaming everything else but the nvidia proprietary linux driver. On x11 the driver literally can’t do anything in parallel without shitting itself to stutters, severe frame drops, judder, and even audio hitches. Just hovering the mouse over buttons, or a popup showing up, or opening an app, or a video playing on browser or mpv and you’re guaranteed to drop frames on e...

                      Remember when you said Wayland had performance issues. Guess what Nvidia has performance issues under X11 as well when KMS mode is used by Nvidia drivers under X11.

                      The Nvidia closed source driver has issues quite major one. Wayland compositor just happens to force KMS/DMABUF usage on. Yes Nvidia X11 driver can at times automatically turn that feature on and let loss hell.

                      The performance issues with Wayland are purely inside Nvidia driver code you can replicate them under X11 using Nvidia closed source driver by forcing particular options on or if you are unlucky enough to have the driver turn those options on by default or because you have used gamescope or other wayland compositor under X11 and nvidia drivers decide to switch over to the KMS code..


                      That the bug degraded performance. In fact the nvidia closed source driver with X11 with right options in fact slows down even more than running wayland compositor
                      It seems to me

                      The Wayland graphics acceleration is provided by Glamor, which has no 3D acceleration capabilities, hence the dire 3D performance compared to X applications and compositors leveraging GLX.

                      That seems to match what I saw benching it 5 months ago.

                      Thats the show stopper on the graphics side of things.

                      On the input side of things, wayland prefers some weird, perverted form of "security" over input latency. That results often in numerous very nasty bugs when that superfluous (because permissions is not a job for a display compositor) layer of security doesnt behave as it should, and best case degraded performance over Xorg-server which doesnt waste time trying decide if an application running on its compositor is allowed to receive keyboard and mouse input.

                      Thats a show stopper for pretty much any 3rd party adoption, and tbh, anyone using wayland as a display server in the first place.

                      Obviously there are a continual stream of issues arising from new application development, X11 or Wayland, but these are not what is holding back 3rd parties adopting wayland, waylands problems are far larger and more deeply embedded in its entire design and delivery.

                      Of course, If i ever think of an application that really needs to not be able to get keyboard and mouse data from a plugged in keyboard and mouse, wayland will be first choice for it, any suggestions?

                      Comment

                      Working...
                      X