Announcement

Collapse
No announcement yet.

Debian Dropping A Number Of Old Linux Drivers Is Angering Vintage Hardware Users

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

  • #71
    Debian is too bloated to run on machines with those GPUs anyway... you'd have to run something really lite like....

    If ConnochaetOS was still around that might be worth throwing on a machine of that vintage. Not too many distros actually strive to support vintage machines.

    Comment


    • #72
      Debian used to be regarded as a clean, stable, quality and freedom oriented distro.
      It looks like it wants to be another Ubuntu now days. It will lose on that path.
      Had it not been for 32bit and legacy support, we would not have Debian on any of the machines we work on or prepare.

      And this is related to the topic of this discussion - we are re-purposing older hardware for use by people who cannot afford computers.
      And there are *plenty of people* like that even in the supposedly rich countries such as the US.

      A 20 year old Pentium-M laptop is still very relevant to someone who has no access to public computers during COVID-19 and needs a modern browser to search for work (we see plenty of such users).
      We do it not because we get paid to do it, but because we want to help those in need.
      We also do not have enough resources to pay those Debian maintainers.

      The privileged approach of some of the overly-compensated (by commercial companies) Debian maintainers, who all of a sudden compare Debian to Windows 10 is appalling.

      To be practical - if Debian is to preserve its user base and reputation - perhaps before one is allowed to be a Debian maintainer she or he need to do a considerable stint as Debian janitors, maintaining this old hardware.

      If Debian abandons its traditions and moves away from that long standing philosophy, it will fade into oblivion, serving as mere jumping board for a few pushy developers to move their careers forward and disappear after a year or two.

      Comment


      • #73
        Originally posted by ssokolow View Post
        You clearly haven't spent any time on retro-tech forums where this sort of thing gets discussed. Here's one answer to someone asking how to do what you describe on the Retrocomputing Stack Exchange.
        One I have spend time doing this.
        The ISA card is a fairly trivial piece of hardware, using I/O ports or memory-mapping only. In this case, it is pretty likely an USB-to-ISA or PCI-to-ISA adapter will work. It is, however, also pretty likely a modern PCI-express or USB replacement is cheaper, thus rendering the adapter useless.
        This point 1 is absolutely true 2000+ ISA old cards I have there is no modern replacement due to be industrial the interface standard until very recently was ISA. Lot of those ISA cards were one offs as in hand soldered on ISA protoboards. Lot of them are full bus control ones.

        The ISA card is not-so-trivial, might use DMA or even take over the bus completely, as many of the more complicated co-processor cards do. I happen to have a Motorola 68040 card on an ISA slot that does take over the bus completely and talks to the graphics card and disk controllers directly - something like that would be worthy of spending quite some of money for an adapter card - I don't, however, know of any adapter card that would support such a setup.
        I deal with stuff like that The usb2isar with ISA daughter board plugged in will handle items like that with a combination of real hardware. usb2 is faster than ISA bus transfer speeds can emulate ISA cards like that. There are issues using the PCI to ISA that you cannot software emulate ISA cards. Basically cards that take over complete bus are common on industrial machinery. Your general cheap PCI to ISA forces you to use daughter board with old cards for the items that take over the complete ISA bus. USB to ISA you can in fact software emulate ISA cards for those kinds of cards. The USB to ISA has mostly replaces PCI to ISA because of the fact you can emulate ISA cards with it to the card in that slot. So a single Motorola 68040 card on a ISA slot in usb2isar slot can be seeing a stack of other cards on the ISA bus with it that don't physical exist just being software generated.

        Please note there are some really expensive PCIe to ISA adaptors that can emulate ISA cards they also have FPGA for doing this so the emulated cards in FPGA on them there is no way I am expecting hobbyist to spend 2000 USD on a adaptor these do find their ways into some industrial usages to make horrible industrial machine controllers work in a smaller space .

        Please note usb2isar used under windows you cannot be doing the fun stuff of software emulated cards. Lot of people who say adaptors don't work for modern hardware issue is they are using Windows and have no clue how to set up the Linux side. Please note using modern hardware does not mean people using adaptors are not using that modern hardware to emulate old hardware so card works and most likely is running really old OS in a virtual machine. So the Linux instance is basically a very fancy hypervisor for operating system setups that only require a 486 this does leave quite a bit of processing room on a raspberrypi or equal to emulate hardware..

        Please note arm items like rasberypi are used alot because you have a really simple start up system like in the uboot class so you don't have this kind of modern system trying to start up old hardware that need hardware emulated before hardware emulation has started. So there is also a split between what you can put in a modern x86 system and what you can put in an adaptor connected to a modern arm system. Of course there are modern x86 systems that support doing custom coreboot or equal so you can make them compatible to run legacy hardware.

        Originally posted by ssokolow View Post
        If you had, I'd expect you or someone you've interacted with to have chimed in with a rebuttal and a link to the adapters you use to solve the problem.
        ssokolow I have no problem with the claim that is not cost effect to use adaptors in a lot cases yes thinking you can end up with some adaptors for some use cases costing more than building a new system from scratch.

        I also have no problem with the claim to set up the adaptors so some of the old cards is complex and a true pain in the ass.
        1) the software setup can complex
        2)Having to pick the right modern motherboard that you can replace firmware on with custom firmware can be a pain as well.
        3)Having to make custom firmware at times is also a mega pain.
        You can start seeing why if the usage case does not require a x86 cpu faster than a arm chip can emulate you end up using a modern arm chip.

        Claiming that you cannot at all is wrong in all the cases I know.

        By the way there is a big reason why you would go to all this effort. Its call reliability and cost of repair and mantaince. Some industrial machine controller going stupid due to some reliability issue could kill someone. This is why software emulation of cards is a option for the industrial usage as this gives less old parts that need to be audited for electrical issues.

        Retro hobbyist has a machine stop working due to something not being reliable normally no one dies so they don't have the motivation in lot of cases to learn how to do all this stuff properly and like the example you just found stupidly claim its impossible so they don't have to consider that path any more.

        Even in the industrial machines the option to redo the complete complete electronics can be cheaper than using adaptor and the man hours required to set it up. Also the fact that industrial usage is for work place health and safety companies get away charging quite a bit for good adaptors so do people implementing them due to how far audited the stuff has to be.

        Comment


        • #74
          Originally posted by 7ray View Post
          Debian used to be regarded as a clean, stable, quality and freedom oriented distro.
          You said stable and quality. The old UMS(User Mode Setting) drivers you really cannot claim to be quality.

          Originally posted by 7ray View Post
          A 20 year old Pentium-M laptop is still very relevant to someone who has no access to public computers during COVID-19 and needs a modern browser to search for work (we see plenty of such users).
          The removed UMS drivers don't in fact make the machine useless. The bios of those machines use vesa in other words the generic Linux kernel modesetting driver will work still. You don't in fact loss the machines means to output screen. You loss the functionality of the GPUs on those cards to accelerate stuff so the machine performs a bit slower. Reality is on those 20 year old gpus you are already normally like 99 percent software rendering the extra 1 percent makes bugger difference for machine useablity when plugged into wall. Also may not operate quite as smooth as it did before but running modern browser it will not be running great the different is most likely less than human will notice. Running off battery you could have a issue.

          Really this a huge storm in tea cup by people who don't understand what the change in fact means. Removing UMS drivers allows kernel to be built without /dev/mem so closing a security flaw for good. Now if enough people want to make that 20 year old hardware work better they need to put time into updating the UMS drivers to KMS(kernel mode setting) or maybe the software rendering mode on that old hardware is good enough as the old hardware will still work this way.

          Comment


          • #75
            Originally posted by oiaohm View Post

            You said stable and quality. The old UMS(User Mode Setting) drivers you really cannot claim to be quality.



            The removed UMS drivers don't in fact make the machine useless. The bios of those machines use vesa in other words the generic Linux kernel modesetting driver will work still. You don't in fact loss the machines means to output screen. You loss the functionality of the GPUs on those cards to accelerate stuff so the machine performs a bit slower. Reality is on those 20 year old gpus you are already normally like 99 percent software rendering the extra 1 percent makes bugger difference for machine useablity when plugged into wall. Also may not operate quite as smooth as it did before but running modern browser it will not be running great the different is most likely less than human will notice. Running off battery you could have a issue.

            Really this a huge storm in tea cup by people who don't understand what the change in fact means. Removing UMS drivers allows kernel to be built without /dev/mem so closing a security flaw for good. Now if enough people want to make that 20 year old hardware work better they need to put time into updating the UMS drivers to KMS(kernel mode setting) or maybe the software rendering mode on that old hardware is good enough as the old hardware will still work this way.
            Reality is that the difference between VESA and card specific drivers on these machines is painfully noticeable in terms of usability.
            Reality is - modern browsers can be used on those old machines.

            Don't know how many of them you use or have to test daily.
            We do it all the time.

            And no, we don't have the resources to translate those UMS drivers to KMS drivers ourselves.

            We are busy enough as it is, trying to make those machines usable for people in need.
            Nor are we against removal of /dev/mem to improve overall security.

            What we are advocating against is - political decisions and arguments that are neglecting the invisible user base - those who really need the legacy support and are not active in expressing their needs in the Debian community, because they are busy with far more real life problems.
            They are a galaxy away from porting these GPU drivers to KMS or fund such development.
            For them a legacy Debian machine is means not a goal.

            This is why the suggestion to nominate candidates for maintainership to be distro janitors first and have them do some legacy clean-up (like UMS to KMS transition).
            This is not a new practice either. Kernel janitors was a good practice years ago.

            Looks like the vast majority of Linux distro developers happily jumped on the Wintel bandwagon, advocating abandonment of perfectly usable older hardware, being overpaid by commercial companies to work on Open Source (check the statistics for yourself).
            They just don't feel the pain of *not* having even $50 to spend on a computer.

            Comment


            • #76
              Originally posted by Djhg2000 View Post
              Out of curiousity, which distro did you move to?
              I've tried many over the years: Suse, Fedora, Netrunner, Slack, KNOPPIX and some ohters. I moved to Ubuntu initially, which was good for a couple of years until I got tired of re-enabling / fixing the graphic stack every six months. At that time I summarized my requirements like this: apt-get, KDE, stability and "not as old as debian-stable". After Canonical made a couple of stupid decisions, I moved on to KDE Neon. But Neon shows that it's not a real distribution. I dropped my requirement of apt-get and and tried something fresh. I was aware of the high praises that Arch usually gets in all the forums, so I tried that. But configuring Arch is a nightmare. I considered MX but decided then for Manjaro. Since almost a year now, I'm happy on Manjaro-KDE.

              Comment


              • #77
                Originally posted by 7ray View Post
                Reality is that the difference between VESA and card specific drivers on these machines is painfully noticeable in terms of usability.
                Ok so you have motivation.

                Originally posted by 7ray View Post
                Don't know how many of them you use or have to test daily.
                We do it all the time.
                That exactly the problem upstream x.org and Debian current debian maintainers have none of the cards left to test with.

                Originally posted by 7ray View Post
                And no, we don't have the resources to translate those UMS drivers to KMS drivers ourselves.
                You have the most important resource to do UMS drivers to KMS the card itself.

                Originally posted by 7ray View Post
                We are busy enough as it is, trying to make those machines usable for people in need.
                That been everyone point of view using this old hardware.

                Originally posted by 7ray View Post
                This is why the suggestion to nominate candidates for maintainership to be distro janitors first and have them do some legacy clean-up (like UMS to KMS transition).
                This is not a new practice either. Kernel janitors was a good practice years ago.
                Kernel Janitors shutdown because there was not stuff simple enough for unfunded newbies do any more.
                https://www.spinics.net/lists/newbies/msg50104.html

                Kernel Janitors program would not have done a UMS to KMS conversion instead would have just discontinued the driver.

                Originally posted by 7ray View Post
                Looks like the vast majority of Linux distro developers happily jumped on the Wintel bandwagon, advocating abandonment of perfectly usable older hardware, being overpaid by commercial companies to work on Open Source (check the statistics for yourself).
                They just don't feel the pain of *not* having even $50 to spend on a computer.
                You are missing the practical realities. Wintel drops hardware at 10 years or less. So that you have 20 years of support is massive. In the last 5 years every one of the UMS drivers being dropped from Debian have had been broken by upstream x.org updates that no one notice before getting to end users because the maintainer and upstream developer at x.org does not have the cards so cannot test what they done.

                Reality is UMS drivers in debian need a new person to step forwards as maintainer with the hardware at the least. The security problem means the drivers need to be migrated to KMS. Again that migration cannot be done without someone with the cards putting in the time testing the changes a developer makes.

                There is no such thing as a free lunch. The reality here is people have wanted less than a 50 dollar computer made from second hand hardware and have been putting zero time do the work to keep the drivers functional on that old hardware.

                By the way developers doing a lot of rebuilds and testing on cards do burn those cards out faster than lightly used hardware. This is what has created the current problem all the cards the developers had are dead and you still have some that still work. Of course once the developer does not have a card to test with he moves on to some other project. Now you are in a complete hole of a problem partly of your own making. Like you did not ear mark any of that hardware to go to developers right it was all for people in need right? You allocated zero people to driver testing to report back to developers if they had broken the drivers.

                This is a domino effect by not keeping maintainers and developers with hardware to test what they are doing they move on this now knowledge base with the hardware reduces and the cost of digging you way out of it increases because now you have to start off with developers and maintainers on X hardware basically at square one.

                See the problem here 7ray its no just the upstream developers who have done the wrong thing. The problem is the upstream developers and maintainers have been left between a rock and hardplace.
                1) keep on providing untested drivers and take flak from users when they don't work this does not work.
                2) cease providing these untested drivers take some flak now and hopefully if there is a enough interest someone with the hardware steps forwards who can do the job properly.

                Parties like you 7ray have done the wrong thing as well so leaving developers high and dry with hardware so unable for the develop newer drivers. This has also left distribution maintainers high and dry for hardware to test what they are doing works as well.

                Remember developers and maintainers can be highly busy they cannot be spending time sorting though 100 old laptops to find 1 that works to test drivers with. 7ray if you are truthful 20 year old hardware that still works right you have scraping a lot of hardware before you get one. Worst part is due to them being at end of life in a month or 2 you can be back looking for another one. The reality is developers and maintainers don't have the time to be doing that sorting process.

                You have been so happy getting distribution maintainers and upstream developers providing free support you have not considered for one minute if they have what they require to-do their job doing that and what was going to happen if they did not keep on getting the resources they need. Yes not getting the resources causes this exact response. Now you are upset at the fall out now. There were alarm bells at the X11 conference 2 years ago when the upstream developers said the had none of this hardware to test with. Not a single person donated a single bit of hardware to test any of the UMS drivers with in the past 2 years.

                Comment


                • #78
                  The hardware is ancient, at this point it would be far more economically-sensible to design and manufacture adapters and a driver (if a simple hardware instruction translation doesn't work) to use low-end semi-modern AMD GPUs in older machines.

                  PCIe->PCI, PCIe->ISA, etc.

                  Obviously the throughput isn't there to start pushing OpenGL4 or Vulkan code, but hardware-accelerated smooth scrolling, video decoding for the UTTERLY INSANE amount of websites with auto loading videos and images (often implemented to COMPLETELY BYPASS settings to NOT download or play them), and displaying system monitors and high-res terminals shouldn't require much.

                  As for other issues old hardware has with modern usage patterns, well,

                  Code:
                  import my_usual_webdesigner_graphic-skill-masturbation_assuming-x86-in-webassembly_and_defaulting-to-hires-data-eating-bs_fuckery_rant
                  Perdone mi linguas.

                  Comment


                  • #79
                    Originally posted by mulenmar View Post
                    The hardware is ancient, at this point it would be far more economically-sensible to design and manufacture adapters and a driver (if a simple hardware instruction translation doesn't work) to use low-end semi-modern AMD GPUs in older machines.
                    Laptops old ones you normally cannot replace the graphics card because its part of motherboard but when you are talking desktop it a different problem.

                    Originally posted by mulenmar View Post
                    PCIe->PCI, PCIe->ISA, etc.
                    There is a practical limit here. ISA is just way too slow to run modern graphics cards stuff will time out.
                    https://www.cdw.ca/product/startech....dapter/2052015

                    A computer with a PCI slot on the other hand it is roughly 100 dollars of adaptors to plug a modern GPU in. Yes card like above to give you a pcie x1 slot and a pciex1 to pciex16 slot for the graphics card to go into. Transfer speed is terrible so you would only use the cheapest entry level pcie gpu even so the performance of this will still be way above the really old gpus for of lot of usages due to video decoding and other things..

                    There are no adaptors cards on the market for AGP to pcie for old systems but I do know it able to be made. The last generation of graphics card for the AGP slot in fact contained a AGP to pcie bridge chip because those cards were 90 percent pcie design. AGP is fast enough in fact to support pcie x4 slot of of the first generation of pcie that would be in more decent range.

                    The reality is a current day pcie video card can be plugged into quite old hardware. PCI bus appeared 27 years ago in 1992 as specification 1993 in real world hardware. Yes that startech pci to pci express adaptor card I just pointed to will work in a 1993 PCI bus driving a 2020 video card in a pciex1 to pcie16 adapter i know because I tried it on a machine where the video card had died and the machine was need up in production as fast as possible to complete a contract. The machine was complete rewired and that old system was scrapped after contract order was complete.

                    VESA Local Bus technically has enough speed for a pcie x1 first generation without time out issues I know of no adaptor so far being made for that and it would only take you back 1 year 1992 in real world hardware..

                    So this is not a case we need to design a lot of adaptors. Design of the adaptors exist for agp and pci slots to pcie cards. pci slot to pcie card is in production. The problem here is affordability is against it. There is also a affordability and supply problem with getting screens that will work with some of the real old video cards as well why it makes sense in industrial to-do this from time to time mostly as short term fixs before replacing most if not all the control system.

                    200 USD in adaptors and graphics card something like a raspberry pi or a smart phone start come possible alternative for the poor at that point. Remember a 200 dollar setup using raspberry pi or smartphone will be way more reliable than a 20 year old computer even with a modern graphics card.

                    Please note all the UMS drivers debian just removed require either PCI or AGP slots for the cards in desktop machines this does mean those motherboards all have PCI slots so the startech adaptor will go in and you can connect a modern GPU to them. The only issue here that is not solvable with money is the laptops of that time frame with the graphics cards built in the motherboard.

                    Really this is another reason why I call this problem mostly a storm in tea cup. Most of the storm is we don't have the money. on one side and developers on the other who don't have the hardware so can no longer make quality drivers for that hardware.

                    Yes those given the poor these old computer that have reliability issues like increased risk of complete hard-drive failure leading to the poor losing data is in some ways not long term helpful. There is a line when computers should go out of usage they are machines they do wear out.
                    Last edited by oiaohm; 23 April 2020, 03:03 PM.

                    Comment


                    • #80
                      Originally posted by oiaohm View Post
                      There is a line when computers should go out of usage they are machines they do wear out.
                      I strongly disagree. I have a working Commodore 64 here that hasn't required maintenance in its entire lifespan. Still boots/runs without issue. Computers can last decades with appropriate care. Sometimes capacitors/voltage regulators need replacement but good hardware lasts for decades.

                      Regarding old hardware support, at a minimum they shouldn't be getting rid of anything that doesn't virtualise properly. QEMU/KVM are good products but they're not good enough yet. There's no Voodoo1/2/3 Emulation for QEMU/KVM yet so it's not possible to GLIDE on Windows 95/98 with accelerated 3D graphics in a Virtual Machine with any sort of useful performance, hell the state of regular 3D performance in Virtual Machines is pretty crap too (no 3D acceleration driver for DirectX/OpenGL on Windows 95/98 under KVM/QEMU either). I know that's slightly off topic since we're talking X11, but there are always funky use cases for old hardware that regular people will not think of. I agree that a Legacy fork of Debian might be required to support these things but nothing should be removed until there's a firm plan in place for a transition.

                      The biggest strength of Linux is the support for things that properietary software refuses to handle anymore. The entire reason I switched to linux 20+ years ago forever was because of Microsoft Sidewinder Force Feedback support being dropped in the Windows 98SE to XP move. Early XP didn't support a $200 product I bought from Microsoft just a year before XP came out. Linux supported it, and even though the Linux support caused a kernel panic when using it, a kernel dev and I fixed the problem and that joystick has been stable on Linux for 20 years. This is part of the magic that makes Linux users fanatical about Linux. It's a key differentiator to Windows and proprietary software.
                      Last edited by DMJC; 23 April 2020, 03:18 PM.

                      Comment

                      Working...
                      X