Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Hardware Linux Technology

Linux Drops Support For 486 and Early Pentium Processors (zdnet.com) 64

An anonymous reader quotes a report from ZDNet: RIP, 486 processor. You've had a long run since Intel released you back in 1989. While Microsoft stopped supporting you with the release of Windows XP in 2001, Linux kept you alive and well for another 20+ years. But all good things must come to an end, and with the forthcoming release of the Linux 6.15 kernel, the 486 and the first Pentium processors will be sunsetted.

Why? Linus Torvalds wrote recently on the Linux Kernel Mailing List (LKML), "I really get the feeling that it's time to leave i486 support behind. There's zero real reason for anybody to waste one second of development effort on this kind of issue." Senior Linux kernel developer Ingo Molnar put Torvalds' remark into context, writing, "In the x86 architecture, we have various complicated hardware emulation facilities on x86-32 to support ancient 32-bit CPUs that very very few people are using with modern kernels. This compatibility glue is sometimes even causing problems that people spend time to resolve, which time could be spent on other things."
"This will be the first time Linux has dropped support for a major chip family since 2012, when Linux stopped supporting the 386 family," notes ZDNet's Steven Vaughan-Nichols. "Moving forward, the minimum supported x86 CPU will now be the original Pentium (P5) or newer, requiring the presence of the Time Stamp Counter (TSC) and the CMPXCHG8B (CX8) instruction. These features are absent in the older 486 and early 586 processors, such as the IDT WinChip and AMD Elan families."

That said, you can continue running Linux on Pentium CPUs, but you'll have to "run museum kernels," as Torvalds pointed out in 2022 when he first floated the idea of ending support for 486.

Linux Drops Support For 486 and Early Pentium Processors

Comments Filter:
  • I get it. Especially considering this is open source. But, how much effort was actually needed to sustain this?

    • Re:I get it. (Score:5, Insightful)

      by Sique ( 173459 ) on Thursday May 08, 2025 @05:47PM (#65362481) Homepage
      A lot. You have to run every test suite for 486. You have to keep a working 486 around for testing. You have bugs to iron out that only appear in 486. You have to run code that's only there because it in software emulates behavior other processors have in hardware. You have to make sure this does not have side effects which don't even appear in other architectures etc.pp..
      • You need to keep not only Intel's 486 models but 486 work-alikes from AMD and others. If a CPU is actively supported all of the variants need to be supported for the claim to be meaningful. Even if that's four different models that still four machines that are over three decades old.

        • by Sique ( 173459 )
          And you have to keep test systems around based on the different chipsets. The list can be expanded: timing is different on different processor architectures. Memory management is different. Interrupt control is different. The number and the layout of differently privileged modes is different. That means that syscalls have to be implemented differently. And this does not in the slightest exhausts the list of 486 quirks and features (and that of every other chip architecture).
    • I get it. Especially considering this is open source. But, how much effort was actually needed to sustain this?

      That's half of the cost/benefit calculation. How much benefit is there? How many people running an operational 486 based system? A "museum display" can run an old kernel.

      If there is an operational 486 system out there that system's devs can update the kernel themselves. The core kernel devs don't need to be involved. It's open source right? At some point it becomes the user's responsibility to make necessary changes or to pay others to make them for them. There is no guarantee that other volunteers will

      • I get that. It's cool that they kept the support this long. I have a PC with a Cyrix MediaGX CPU, which is newer than a 486 and still the newest I can run on it is Debian 8, anything newer and the setup does not even start (not enough RAM or the CPU does not support some feature). I have not tried to compile the kernel or anything, I'm just using Debian 8 on that system.
        I have a PC with 486, but would not think to try to run the latest kernel on it.

        • by drnb ( 2434720 )

          I have a PC with 486, but would not think to try to run the latest kernel on it.

          Do you still have a 486, have your turned it on lately? I turned a Pentium on for the first time in years and there was a pop followed by a burning smell. May your motherboard's capacitors be better than mine. :-)

          • Yeah, last year. I installed Windows 3.1 on it and use it sometimes to play old games.
            Bad caps are not difficult to replace, but I did not need to do it on that particular PC.

            I also have a motherboard with a 286 (last tried to use it a few years ago), but do not have a suitable case for it.

  • by Anonymous Coward

    It's not even true that you'll "have to run museum kernels" with no security support if you've got affected hardware.

    There are LTS kernels from kernel.org with at least two years support from now, and distro kernels prior to V6.15 which will be supported with backported security fixes for 10+ years.

  • by Entrope ( 68843 ) on Thursday May 08, 2025 @05:52PM (#65362491) Homepage

    The Linux team is finally removing support for a 36-year-old processor ... and still supporting a 32-year-old processor. Meanwhile, the commercial OS people are over there killing support after seven or ten years. That shorter lifetime has not been necessarily bad for enterprise desktops in rich economise ... but an awful lot of the world falls outside that category. On top of that, Moore's law no longer delivers nearly as much year-over-year gain in performance, so it makes sense to keep hardware for longer and longer periods.

    • by ledow ( 319597 )

      If the world ever ends, it ain't gonna be old versions of Windows that you'll be booting up.

    • by _merlin ( 160982 )

      Linux has removed support for processors a lot faster than that. Support for DSPs often doen't last very long, and Tilera TILE64 was possibly the architecture with the shortest-lived Linux kernel support, being added in 2.6.3 (2010) and removed in 4.16 (2018).

      • by Entrope ( 68843 )

        How long did Windows and MacOS those architectures?

        If you're going to make comparisons, they should hold up to scrutiny.

    • Your post is non-sequitur. Moore's law slowing down now does not in any way have an impact on a 36 year old processor which was itself virtually obsolete a year or two after it came out.

      And you're ignoring capability and purpose of computer hardware.

  • by Valgrus Thunderaxe ( 8769977 ) on Thursday May 08, 2025 @05:57PM (#65362497)
    But I'm not.

    Look at all the junk ISA and Microchannel nonsense in the kernel. Get it out of there - and if these modules are actually maintained, then they can go up on GitHub and if people want this, they can get it there.
    • by vadim_t ( 324782 ) on Thursday May 08, 2025 @06:03PM (#65362505) Homepage

      Amazingly, ISA disappeared much later than I thought. It seems there's a Skylake motherboard with an ISA slot out there. That architecture was only discontinued by intel 6 years ago.

      Also, I believe the bus still internally survives in many boards that don't have a physical connector, and the LPC bus is ISA in a slightly newer form.

      • Re: (Score:2, Troll)

        by drinkypoo ( 153816 )

        It's very common for sensors to be on ISA, but that's really about it any more. It was common for a long time for mouse and keyboard, but I don't think even they are there now. If you put the sensors on some other bus, and there are several options, you could reasonably ditch ISA.

        Hmm, wow, apropos of nothing except running sensors to see whether I was still using ISA, I seem to have a bunch of temp sensors in my system now. Besides the mainboard (which uses a nct6798, which it says is on ISA) there's the tw

      • by AmiMoJo ( 196126 )

        There are chips that do PCIe to ISA and USB to ISA. You can get adapters from Chinese manufacturers who are still supporting industrial users.

        They still make 8086 machines and everything in-between over there. With such a huge global market, nothing is too niche for them. Need a replacement laser assembly for your Laserdisc player? An IDE SSD? Token ring USB adapter?

      • by smash ( 1351 )
        Like it or not there are legacy industry specific hardware adapters out there for ISA. That's why it was quite late to be removed from motherboards.
    • by WhoBeDaPlaya ( 984958 ) on Thursday May 08, 2025 @09:52PM (#65362911) Homepage

      You might want to look into LPC bus - ISA is still alive and well!
      https://www.youtube.com/watch?... [youtube.com]

  • I remember the flame wars in the 80's about how the 486 shouldn't even be in the release kernel.
    And how the 486 will outlive the free linux kernel.

    • Re: (Score:3, Interesting)

      by Anonymous Coward

      Not sure if this is joke or not, but the Linux kernel came out in 1991.

      At the time, nobody did anything long with releases. At best, you had 80x86, which went back to 8088, MS/DOS and CP/M. It is still pretty amazing how long compatibility was kept since then.

  • by manu0601 ( 2221348 ) on Thursday May 08, 2025 @06:13PM (#65362521)
    NetBSD still supports 386 and newer. However such ancient system often has old IDE controller that are weak at asynchronous I/O, that makes the experience very sluggish. With an SCSI drive, that gets better.
    • by jmccue ( 834797 )

      Came here to say the same, and AFAIK there are no planns for NetBSD to drop that support.

      OpenBSD is there too, but seems the day may come for i386 (32 bit) removal, see:

      https://www.openbsd.org/i386.html

      Due to the increased usage of OpenBSD/amd64, as well as the age and practicality of most i386 hardware, only easy and critical security fixes are backported to i386.

      • i386 (mind the leading i) is the NetBSD and OpenBSD name for the port to any x86 32-bit CPU. Hence dropping i386 support does not just mean dropping 80386, but also 80386, pentium, pentium pro, and clones.
  • But will he kill off the transmeta crusoe, his own child?
  • by 4wdloop ( 1031398 ) on Thursday May 08, 2025 @06:19PM (#65362539)

    So is it only ARMv7, 586 and 686 left in 32-bit CPUs?

  • The ZF Micro "ZFx86" SoC lacks the needed instructions and thus will no longer be able to run new versions of Linux.
    These are used in embedded systems, so impact is likely to be on some industrial controllers.

    On the up side, the Vortex86 line is more common and will continue to function fine with new builds.

    • Oddly enough, the software already running on those controllers won't simply disappear, and the last working build will also still be available.
      • The problem is that, CVEs that arises in the kernel are unlikely to get patched for these systems which is problematic because these are used in embedded industrial systems. The result is that these systems are likely to remain vulnerable because replacing them is a pain in the ass and expensive. So, critical infrastructure is now potentially at risk from a future nation-state attack.

        • If your embedded industrial systems are accessible by 'nation states,' the state of your patching isn't the issue, nor is it going to prevent exploits.
  • Instead of wasting human hours manually fixing the "compatibility glue" they should have written and trained AI to deeply understand all glue from the past and present and then used it to automatically keep the glue updated for all future versions without wasting human time.

    These fools are stuck in the past. There is no reason not to support i486 forever.

    • by Anonymous Coward

      You forgot the /s.

      (I hope.)

  • The lower end of Vortex's line of enbeded processors has processors that will be out of support after this. Lucky for them, the Linux for Public infrastructure will maintain some LTS for even longer (~ ten years)

    Is the same reason that ISA, Microchannel, EISA, VESA and OG PCI support is still ahnging around in the kernel. Many PC/104 systems still being built today, some of those, with that stuff

    Lucky for the X86-32 embeded world, the upper end of the Vortex range, and many Via processors still made today a

    • by tlhIngan ( 30335 )

      And those Vortex processors support some rather modern hardware - they have SD card controllers that simulate a SATA host adapter (backwards compatible to ATA) so you can run a legacy OS without giving up modern easier to get hardware. Likewise they have DDR4 and USB support. They also have PCI, ISA and PCIe bus support.

      They just can't implement the Pentium instruction set due to patents so they have a 486 core running at up to 1GHz or so.

      They are extremely popular CPUs for embedded PCs - the screens you se

      • by Anonymous Coward

        Any Pentium patents expired more than ten years ago.

  • You bastards! (Score:2, Insightful)

    by Randseed ( 132501 )
    I'm still running an 80486DX 50MHz with 16MB of RAM, you insensitive clod!
  • The Slashdot editor who posted this story obviously is confused about which CPUs are "Pentiums" and which are not.

  • Even this is not enough, drop all support for 32-bit x86 and make the new minimum target amd64.

  • You can still find and install older kernels that do support older CPUs, if you really need them.
    • You can still find and install older kernels that do support older CPUs, if you really need them.

      This is kind of a nothing-burger. Anything running a processor that old is going to have hardware that is correspondingly old and probable be a niche case anyway, so you could just go use the latest kernel that supports the hardware involved. It will run as well as it did back then. Of course, you lose out on a lot of improvements, but that is to be expected, and nobody is targeting the mainstream kernel to be optimized to that hardware, so nothing is lost.

    • Also Linux is open source; if someone wants to maintain 486 support in their kernel version, they can. Linus will no longer maintain it as part of mainline.
  • You'd think with something as momentous as dropping a (the?) major CPU arch of 486, they'd bump to 7.00 instead of putting it in a point release. This is certainly a breaking change.

The reason why worry kills more people than work is that more people worry than work.

Working...