Linux Nukes 386 Support 464
sfcrazy writes with news that Linus pulled a patch by Ingo Molnar to remove support for the 386 from the kernel. From Ingo's commit log: "Unfortunately there's a nostalgic cost: your old original 386 DX33 system from early 1991 won't be able to boot modern Linux kernels anymore. Sniff."
Linus adds: "I'm not sentimental. Good riddance."
Dammit (Score:5, Funny)
I'm forking Linux right now to support this under-appreciated processor.
Re:Dammit (Score:5, Insightful)
If you HAVE a 386, don't you also REALLY want a pre-2.0 kernel, anyway? :-)
Re:Dammit (Score:5, Funny)
If you HAVE a 386, don't you also REALLY want a pre-2.0 kernel, anyway? :-)
Real men run .99, and wait 16 hours for their kernel to compile.
(Of course, that was 1/4 the time it took X to compile ...)
Re: (Score:3)
Real men run .99
No, those sort of people are "children". When you mature you stop using the term "real men".
Re:Dammit (Score:4, Funny)
...or a spinoff from a band that sucks.
(for the clueless: I have their entire discography.)
Comment removed (Score:5, Insightful)
Re:Dammit (Score:5, Informative)
And if you don't require 386 support frankly a $25 ARM thumbstick will give you much more work per watt while being even lower powered than the Bobcat or Atom.
Even if you do, your ARM thumbstick can probably emulate the 386 instruction set at a faster rate than the original chips, via Bochs or QEMU.
--Joe
Re: (Score:3)
Re: (Score:3)
I disagree. Dumpster diving has never been better. Nowadays, pretty much any typical dumpster find is going to a be a viable computer in the sense that it will do perfectly fine for what most people would want a computer for (assuming it works, of course - a good number of succumbed to the capacitor plague and it's not worth trying to repair an otherwise uninteresting P4 system). Perfectly capable of running the latest Linux distros (sorry no 386's) or Windows XP. Vista or Windows 7 isn't even out of th
Re: (Score:3, Funny)
Good luck getting Windows 8 to run on your 386.
Re: (Score:3)
What about those of us who are still using Windows 3.1? On a 286.
Re: (Score:3)
Re:Dammit (Score:4, Insightful)
I have seen a modern embedded x86-compatible CPU platform with 256MB of RAM. The CPU was not by Via, but it was an x86 compatible one that implemented only the 486 instruction set.
Great fun to be had as most of the Linux stuff we had was compiled for i586 and higher. It just crashes.
Re: (Score:3)
Re: (Score:3, Informative)
Yes. Sequent Symmetry S81, up to 384MB of RAM, and up to 30 80386 processors.
No, it won't fit under your desk.
Re:Dammit (Score:4, Interesting)
Re: (Score:3)
30 pin SIMMS are readily available at least up to 16MB a stick, they're used frequently in the vintage Mac world to get the likes of SE/30s and IIcis up to 128 MB of RAM. I can't say they're the biggest but it's the biggest I've seen.
Here [macsales.com]
They're really surprisingly affordable for what they are. I suspect they'd work fine in the majority of 386 boards (and I've seen 386 boards with 8 SIMM slots)
Comment removed (Score:5, Informative)
Re: (Score:3)
Re: (Score:3, Funny)
Why? 640 KB should be enough for anybody.
Re: (Score:3)
Are you sure that wasn't 1MB not 1GB? I think the biggest memory sticks that existed back then were 1MB... you'd need 1024 of them to make a GB.
Re: (Score:3)
Yeah, I think he must be three orders of magnitude off.
My earliest computers include: ... and my main mail/nis/nfs/dns/dhcp/web/samba server in 2012 has 512 MB and runs quite well on a single core.
Sinclair ZX-81: 1 kB (upgraded to 16 kB)
Commodore 64: 64 kB (38911 bytes usable)
Sharp MZ-700: 64 kB (48 kB usable)
Amiga 1000: 256 kB (upgraded to 512 kB)
Homebrew PC, 386 turbo: 640 kB
Amiga 4000: 4 MB (2 MB chip ram, 2 MB fast ram)
Heck, my old A1000 running from floppies was more responsive than my new laptop
Re:Dammit (Score:5, Interesting)
I worked for Micro in the early 90's. I loaded up a 386 with 256MB of RAM. We made expansion slots that fit into the ISA slots. I filled every ISA slot with a fully loaded expansion card.
When the PC booted, I had the AutoExec.BAT create a Huge RAM drive, then copy the contents of the Windows directory to the RAM drive and launch Windows from the RAM drive. When we shutdown, we ran a batch file that copied the RAM drive back to the Hard drive.
It was the fastest Windows 3.1 system in the company.
First time I ran a computer with a Flash Drive on it, it felt like that old system.
Re:Dammit (Score:5, Insightful)
Before 1990, I was using a 386 with 4MB of ram. In 1991, my parents purchased a 486sx 25mhz with 16MB of ram for $1500. The hard drive was 170mb. If you had a gig of ram, why even need a hard drive? You must have just created ramdisks and had a blazing fast computer.
By 1997, I had colocated my first server on a pentium dual xeon 450mhz with 512mb of ram. This system cost upwards of $2k to build at the time.
I'm almost sure 386 had NO support for dimms. So you used simms? Were they 30 or 72 pin?
From http://en.wikipedia.org/wiki/SIMM [wikipedia.org], 72pin simms did not replace 30pin simms until the mid 90's and were NOT present in 386's. 30pin simm sizes ranged from 256kb to 16mb while 72pin simm sizes ranged from 1mb to 128mb.
Before spamming us with your nostalgia, at least try to get your facts within a magnitude of the truth.
Re: (Score:3)
Bullshit. There's no way 16MB of RAM plus a whole 486-based computer would have gone for as little as $1500 in 1991. You could get a Pentium-based machine around 1995 with 16MB for in the neighborhood of 3 grand.
Re: (Score:3)
Your 386/20 had 24 address pins, ergo it could address 16 MiB of RAM.
Re: (Score:3)
The 386 chips were most commonly paired with 1MB RAM. :-)
While you're correct, my first Linux machine was a 386DX25 with a full-AT motherboard, which I got in a case with a power supply and a floppy drive. Much of that space was used for 8MB of DIP DRAM (fully populated.) No idea who made the board. I added a 120MB Maxtor, 1MB Trident VGA card (ISA of course) and Slackware 2.0.
Re:Dammit (Score:4, Informative)
Re: (Score:3, Funny)
You would have a really hard time maintaining it. The stuff that was removed allows them to change a whole mess of things that will become incompatible to backport from
People still using 386's probably don't update very often.
Re:Dammit (Score:5, Informative)
I'm honestly surprised that it held on this long.
Intel EOLed even their embedded 386s sufficiently long ago that I had to go to archive.org to find the discontinuation notice [archive.org]. The last 386 rolled out the door in 2007.
There still seem to be some other outfits I've never heard of making x86s for embedded applications, but the specs on those boards are sufficiently primitive that they generally seem to be aiming for DOS, not the leading edge of the 3.X kernel tree.
Re:Dammit (Score:5, Informative)
Debian dropped 386 support way back when 3.1 came out. Here are the reasons. [debian.org]
Time to fork (Score:5, Funny)
I'll fork the kernal and keep the 386 dream alive, just as soon as the checkout is complete on my blazing fast 28.8k modem.
Re:Time to fork (Score:4, Funny)
Re:Time to fork (Score:5, Informative)
Debian dropped i386 kernel images a very long time ago; the lowest you can go is 486.
Annoying for me is, that they also dropped i686 without pae. Meaning for my AMD Geodes I either have to roll my own or install 486.
Re: (Score:3, Funny)
Oh man - another show-off!
Look - I don't mind if you are happy with your shiny super high speed modem, but will you keep your tone a bit down and feel a bit for us oldies that are still on a Bell 101 100 baud modem will you?
Incredible how far someone will go to peek both fingers in the eyes of some less wealthy helpless underdogs... :-(
Fond Memories (Score:5, Funny)
I remember furiously masturbating to low resolution pornography on my 386.
640x480 is perfectly wankable if you ask me...
Which nobody does.
Why don't you call me anymore??
Re:Fond Memories (Score:5, Funny)
Re:Fond Memories (Score:5, Funny)
You kids and your artificial porn. Back in my day, we had to have sex with real women. And we liked it!
Re:Fond Memories (Score:5, Funny)
ASCII porn on the line printer.
Re:Fond Memories (Score:5, Funny)
ASCII porn on the line printer.
Don't knock it, man... it was good practice for not being a two pump chump with the ladies. When it takes 15 minutes of watching the paper scroll to get to the good part, you learned to take your time ...
Re: (Score:3)
Re:Fond Memories (Score:5, Informative)
I never knew anyone who regularly made the distinction between "line printer" and "dot matrix printer" when talking about "line printer ASCII art". Sure, line printers were their own thing, but when used as an adjective, it was always synonymous with DMP. Now get off my lawn, or I'll rant about how ttys are actual teletype machines, and not just a damned serial port!
Re:Fond Memories (Score:5, Funny)
ASCII? An extravagance. Real men had an Altair 8800 that flashed its lights suggestively.
Re: (Score:3)
What was the last version which actually did? (Score:4, Insightful)
Re:What was the last version which actually did? (Score:4, Insightful)
Re:What was the last version which actually did? (Score:5, Informative)
Re: (Score:2)
So, how many 386 computers actually have enough RAM to handle a modern kernel?
The size of the kernel had certainly bloated with module & stuff. Yes, I am aware that you COULD custom-compile a kernel with just what you need, but would that even fit in 32Mb RAM (probably a pretty good amount of memory when the 386 was king)?
Re: (Score:2)
My 386 has 4mb of RAM and no math coprocessor. No Linux on this thing.
Re: (Score:3, Funny)
I had a 386/40 with 8MB and got it to run two instances of Quake with heavy swapping in Linux. I think it even got 30 seconds per frame since it was using an emulated math coprocessor.
Re:What was the last version which actually did? (Score:5, Funny)
30 seconds per frame
I see what you did there...
Re: (Score:3)
32MB? That was a pretty good amount of memory when Pentium was king.
Re: (Score:3)
Bearded UNIX admin: (Score:5, Funny)
Now I Want... (Score:3)
Re: (Score:2)
Re: (Score:3)
Click here if you want your mind blown. [sourceforge.net]
Re: (Score:3)
With DOS you can actually IRC from an original IBM 5150 PC. Mike Brutman's mTCP includes an IRC client that runs on any PC. The only thing that keeps me from using it regularly is the lack of multi-server support. mTCP also has an FTP client and FTP server (!) which have become my favorite means for putting files on these old PCs. Far more convenient than zmodem.
Oh well. (Score:3)
Re: (Score:2)
On noes! The satellites! (Score:2)
If I remember correctly, don't quite a few sats run on i386 and i486 procs? They'll fall from the skies in protest! Oh noes!!
Realistically speaking, I'm kind of glad to see it go. Especially if they have been having to make things overly complex trying to retain backwards compatibility.
Re:On noes! The satellites! (Score:5, Insightful)
Especially if they have been having to make things overly complex trying to retain backwards compatibility.
Now, see... if he'd just gone and written a microkernel in ther first place, we could support multiple processor architectures with a single codetree anyway....
Re:On noes! The satellites! (Score:5, Funny)
Tanenbaum, is that you?
Re: (Score:3)
Well, Linus went with a monolithic kernel, but others already HAVE made a microkernel. Every heard of Hurd? I understand that, after only 20 years of development, it will go stable any year now.
Re: (Score:3)
I was just trying to set the cat among the pigeons, so to speak.
But now that we're on the subject: it took a monolithic kernel for the project to be manageable to a single coder, and Linus made the original kernel himself. The Hurd isn't finished because a decent microkernel needs a lot of development time. Paradoxical, then, that the Linux community has a critical mass of development talent that could knock together a microkernel achitecture in a couple of months, if they wanted to, but the projects with
Re: (Score:3)
Nope. The point of a microkernel is about long-term stability and code maintainability, and run-time security. User-space services shouldn't be able crash the kernel, so the kernel can just restart the services. There are some elements of this in all *n*x systems -- print daemons and the like.
The Linux monolithic kernal was easier to get up and running than the Hurd microkernel, but there's a lot of legacy stuff drifting round the Linux codebase because a lot of Linux's development was hack upon hack.
No point supporting i386 anyway (Score:5, Informative)
No point supporting i386 anyway
As far as I'm aware, GNU binutils won't work on anything less than a 486
I guess you could be affected if you're using some other toolchain, but realistically is pointless keeping support for 386
Nostalgia (Score:2)
My first server was a 386sx16 with 4MB of memory - sad Gothmolly is sad.
Silicon (Score:3)
By the way, I believe that the 387 math coprocessor has been axed on the 64-bit side of the processors and SSE2 is explicitly used for mathematics there instead.
Going further, I wonder if it is possible to rip the 32-bit parts completely away from the silicon at some point?
Re: (Score:2)
Do you want that to happen before or after ripping out the 16-bit parts? Even the latest 64-bit CPUs boot up in 16-bit mode. As far as I recall you still need 32-bit mode because there isn't support for switching directly from 16-bit mode to 64-bit mode.
Are there any AMD64 (or compatible) CPUs, which can be powered on directly in 64-bit mode? Supporting that would be the first step towards get
Re: (Score:2)
Re:Rad-hardened processors? (Score:5, Insightful)
There are. But how many of them desperately need to run a 3.7 kernel?
Re: (Score:3)
They don't need a bleeding edge kernel.
On that kind of hardware you would be running 2.4 or older.
That terminal emulator... (Score:2)
Historical significance (Score:5, Interesting)
The historical significance of this of course is that Linux was originally written to specifically target the 80386, and it was written with the 386 with *no* portability in mind. So it no longer supports the CPU it was originally written for.
Re: (Score:2)
Re: (Score:3)
Anyone in the world affected at all? (Score:4, Insightful)
I'm trying to figure out if any user, worldwide, would be affected by this.
As pointed out in another comment, there aren't very many applications that will work. If anyone, worldwide, is using it as a desktop OS, they probably are on an older kernel anyway.
As for embedded systems : since new 386 CPUs have not been produced in 5 years, there's not anyone who would be designing a new embedded system that will use a recent kernel. There's old systems deployed in the field - but why would anyone try to upgrade an old embedded system to a new OS and kernel? A good embedded system is supposed to be reliable and simple enough it needs only minor bug fixes throughout it's deployed lifespan.
Re:Anyone in the world affected at all? (Score:5, Informative)
Re:Anyone in the world affected at all? (Score:5, Interesting)
Just 386? (Score:2)
What is the new absolute minimum spec for Linux?
How old can you go? (Score:3)
What's the oldest system anyone has that's still in use?
I've got a Dell Dimension XPS Pro 200n that's been going nearly 24/7 since the late 90s, shut down only to move locations. It hosts a Citadel BBS for a small group of old timers. I replaced the hard drive last year when it started making alarming noises and crashing randomly but everything else is original. Some day soon, I'm going to virtualize it and find a cheap host. Of course, I've been saying that for over a year.
Re:How old can you go? (Score:5, Interesting)
But would you really want to? (Score:3)
Considering the limited resources of such old hardware and resource requirements of newer software.... it is better to stay with lightweight older versions of Linux or other OS's to keep these systems in use. One such OS might be AROS.
WTF? English fail (Score:5, Informative)
At first I thought I was going crazy. If Linux "pulled a patch by Igno" to remove 386 support, then that would mean that he prevented the patch going in. So why does he add "Good riddance" at the bottom?
So then I read the second link [muktware.com] and it actually says:
I've been here a while and this is the first time I can remember that I've seen a story on Slashdot state the complete opposite of what actually happened. Geeeeesh.
Re: (Score:3)
"git pull".
Learn it.
Re:WTF? English fail (Score:5, Informative)
To "pull" a patch is a git-ism because you use the command "git pull" to bring in changes from remote repositories.
Re:WTF? English fail (Score:5, Informative)
In kernel speak, "pulling a patch" means he accepted someone's patch.
I love how every profession or hobby introduces it's own jargon.
Re: (Score:3)
In Slashdot speak, "it's" means "its". :D
I love how every editor or user introduces its own mistake.
Look at the patch. (Score:5, Interesting)
Sometimes, it seems these "removal" patches are more for religious reasons (aka break it on purpose) than any kind of technical ones. Same thing when firefox removed PPC or windows 2k support.
In fact I bet if you compiled a non SMP linux kernel it probably still works (assuming it does actually still work on a 486/pentium), as the majority of the patches are related to CAS and page invalidation, which aren't really necessary anyway.
ought to change the kernel version number to 4 (Score:4, Insightful)
This is a big enough change to warrant version 4.0. Otherwise, we might reach kernel version 3.8.6 which won't work on a 386.
What I don't understand is what change between the 386 and 486 makes dropping the 386 a good idea. What functionality has the 486 got that the 386 doesn't have?
Re:ought to change the kernel version number to 4 (Score:5, Informative)
good run, good backward compatibility (Score:3)
Kudos to linux for ditching it only when it became a pain to maintain co-compatibility with newer stuff.
Instead of yanking it preemptively to cattle-prod people onto the upgrade treadmill.
Hello. Your name is Ingo Molnar. (Score:4, Funny)
You killed my CPU. Prepare to die. [imdb.com]
Ingo rants is bliss (Score:2)
Re: (Score:2)
You killed my father...
Re:So... (Score:5, Informative)
Re: (Score:3)
80 is perhaps too restrictive, but the limit shouldn't be much higher. One problem is that the distribution of line width (and thus overall code density) in any sufficiently long file can vary erratically and this makes for inefficient consumption by a human reader.
Also, the power of parallel editing should not be overlooked! Sometimes it's nice to be looking at 4 source modules simultaneously.
Re: (Score:3)
Re: (Score:3)
I don't think there was really a 66mhz 386. The fastest I've seen is 40mhz. IIRC, the 486dx2/66 was the first clock doubled x86 processor which ran at a 33mhz bus speed. So if a 66mhz 386 existed it would have to run on a 66mhz board. Did those exist?
Also, the 100mhz 486s were DX4s. That's 33mhz bus * 3. Intel skipped DX3 for trademark reasons. Of course, you might have a 100mhz DX2 if you ran the board at 50mhz. You can do it with a Cyrix chip [alasir.com]. I don't know if you can do it with an Intel DX2.