BSDs and my AMDK6-2 400 rig with 256megs of ram.

[Eudimorphodon]

Now that is useful information.

I don't know what exactly you want from this thread, so...?

FWIW, you might have some options to *try* fix the video corruption on the Net/OpenBSD machines. Long story short: Modern X.org doesn't run with a set configuration file, it autoconfigs everything. It doesn't always guess everything right, especially on old obscure video cards or unusual OSes. All I can suggest is you can try RTFM-ing about how to generate an xorg.conf configuration file and seeing if you can tweak some of the options:

https://www.x.org/releases/current/doc/man/man5/xorg.conf.5.xhtml

Poking around various mailing lists it seems like over the years a lot of people (including on Linux) have had mixed luck with Matrox cards. The first thing I would suggest is:

Option "Accel" "Off"

And see if the corruption goes away. Obviously that's not optimal, but it would be "interesting". Sometimes turning off hardware mouse cursors magically fixes random problems as well.

 

[Caluser2000]

It was just a post on WHAT HAPPENED on my AMDK6_2 400 rig in respect to installing those BSDs for goodness and that was ALL it was about.


How is that hard to understand?

I'm not worried about swapping cards at all and as I replied to Chuck I am going to give them a go on my P3 800 system out in the shed. If if doesn't work out I'll try other video cards and make a post about it.


I truly hope haven't wasted any time seeing if there is a solution.....We'd already decided the video card was the problem for OpenBSD and NetBSD. I'm not sure what's up with MidnightBSD or FreeBSD from which it is derive dfrom-ie same problem at the same place -ie fail completely to even start the installation routine. I do not even intend to discuss that in this thread at all and will not reply to ANY posts regarding that. It is a dead issue and was a complete fail , on differentt media no less, on the on my AMDK6-2 400 system. Maybe when I make a post about the results on the P3 800, if they are successful. might even try on my AIO Compaq Presario 5522 from 1995 I'm upgading and modify to have a nic put in the place where the 9600 dial up modem was. It has 64megs of ram so one of the BSDs should be able to handle that. But I'm certainly in no rush to do those. Other projects have a higher priority at this point in time. Like getting my XTRA Professional Series XT Turbo up and running.

I can post whatever I like here because it is the *nix related section.

 

[tokenalt]

To put it another way, in the grand scheme of things that NetBSD can still "run" on something as limited as a Sun2 workstation is primarily useful as a demonstration of how it's possible to trim the OS down enough to fit in a teeny-tiny box, it's not really a very useful thing in and of itself. (How many people do you know that still have a Sun 2/50 in working condition and no better options to do... just about anything, on?) This means that even within the Tier1 ports if there's some hardware device with a driver that's troublesome because it's rare and has no one willing to support it it's fair game to get deleted, same as it is in Linux. I think the BSDs do it less often in part because their focus on device independence and slower evolution has meant that there is less device-breaking "churn" in the kernel interfaces, but ultimately if nobody is using something support might still go someday even though the OS continues to claim support for machines of that "era".

Funny you should mention a 2/50 as I've actually been net booting netbsd on a 3/80 lately and it works out of the box but is extremely slow. First boot was 6h40m long, second boot was only 30m long. It's actually pretty amazing the difference in performance I get from this machine vs. a sparcstation 10 from just a few years later which boots up in a minute or two.

As for the lack of video card support you can blame xorg for that as they dropped support for almost all 2d cards a few years back. NetBSD has maintained support for most non x86 video cards but there doesn't seem to be much interest in old x86 hardware, probably cause most of the people that play with retro x86 run dos and windows.

 

[Caluser2000]

Linux/X.org runs just fine with the card .

The *BSD don't out of the box.

The big question is WHY and I've mentioned it a number of times. What are they using as source code to implement Xwindows its drivers on BSD that is different from Linux?

Does no one except Chuck actually read my post?

 

[Caluser2000]

If you'll want a thread on system the run all functions of BSD fine and the "tweaks" to sort out issues I'll create on if you like?

It'd be interesting and usefull to other members I should imagine.

 

[Eudimorphodon]

The big question is WHY and I've mentioned it a number of times. What are they using as source code to implement Xwindows its drivers on BSD that is different from Linux?

Does no one except Chuck actually read my post?

Funny you keep asking that, because it’s pretty clear your reading something into the responses that simply isn’t there. Or at least I’m going to kindly go with that interpretation, because otherwise the level of hostility you’ve expressed here is frankly baffling.

It’s been pointed out several times that:

A: you are not making an Apples to apples comparison when you pit 2016 Linux against 2021 BSD; the way the BSD package source tree works isn’t the same as how Linux distributions handle code versioning, so it is in fact very unlikely you’re dealing with the same versions of code even if your Slackware is “up to date”.

B: they’re different OSes, there are going to be edge cases different between the two, and:

C: It could be a minor issue that could be manually tweaked, but you bit my head off for suggesting you lift a finger trying to figure it out.

So, whatever dude.

 

[whartung]

Funny you should mention a 2/50 as I've actually been net booting netbsd on a 3/80 lately and it works out of the box but is extremely slow.

Is this a "period appropriate" NetBSD? Or a recent version?

First boot was 6h40m long, second boot was only 30m long.

Why would the first one take so long? 30m is still "unusable" in my book.

We used to run Sun 2/50's back in the day, and, yea, they're as bottom of the barrel as you can get (I think they're just 16MHz, if I recall), but they were usable. We ran SunWindows on them (you had too, the default Sun console is just glacial), and it ran fine. Mind, we did not net boot them, they had a local drive. But for the day, it was a usable machine.

 

[Caluser2000]

Funny you keep asking that, because it’s pretty clear your reading something into the responses that simply isn’t there. Or at least I’m going to kindly go with that interpretation, because otherwise the level of hostility you’ve expressed here is frankly baffling.

It’s been pointed out several times that:

A: you are not making an Apples to apples comparison when you pit 2016 Linux against 2021 BSD; the way the BSD package source tree works isn’t the same as how Linux distributions handle code versioning, so it is in fact very unlikely you’re dealing with the same versions of code even if your Slackware is “up to date”.

B: they’re different OSes, there are going to be edge cases different between the two, and:

C: It could be a minor issue that could be manually tweaked, but you bit my head off for suggesting you lift a finger trying to figure it out.

So, whatever dude.

Did I mention Slackware 14.2 was still supported. In other words UP TO DATE. You seem to be missing that little snippit lol.

Slackware 15 has just been recently released.

I'm sorry to see you have taken this so seriously. I really really am.

 

[tokenalt]

Is this a "period appropriate" NetBSD? Or a recent version?

Why would the first one take so long? 30m is still "unusable" in my book.

We used to run Sun 2/50's back in the day, and, yea, they're as bottom of the barrel as you can get (I think they're just 16MHz, if I recall), but they were usable. We ran SunWindows on them (you had too, the default Sun console is just glacial), and it ran fine. Mind, we did not net boot them, they had a local drive. But for the day, it was a usable machine.

NetBSD 9.2 and the long boot time was mostly from generating the fontconfig cache which was about 6 hours. Even after booting the system is barely usable because of makemandb running in the background for several hours. I think a large part of the slowness is due to running everything over nfs so anytime the system does anything it's flooded with interrupts, top shows a constant 20-40% interrupt time. I'll try installing on a hard drive and hopefully that will improve performance.

 

[boomer3200]

You can add makemandb=NO and fccache=NO in /etc/rc.conf before first boot.

 

[whartung]

NetBSD 9.2 and the long boot time was mostly from generating the fontconfig cache which was about 6 hours.

Yea back then the fonts were chiseled by hand out of blocks of bits, nothing to cache.

Even after booting the system is barely usable because of makemandb running in the background for several hours.

Another one of those modern systems that we take for granted because the excess bandwidth to do such things in the background and not notice it, not to mention having the disk space to burn to feed the eternal "memory/storage vs performance" tradeoff.

Back in the day we relied on "man -k", and I honestly have no idea when that db was built -- probably just installed with the stock man pages.

Yea, "makemandb" is the latest generation for man -k (and apropos), and uses a SQLite db for its index (which, of course, make a whole lot of sense today). But certainly a more expensive option back in the day.

And, yea, your NFS may well be holding you back. I honestly don't recall if we were using NFS back with our Sun 2/50s.

 

[Eudimorphodon]

I think a large part of the slowness is due to running everything over nfs so anytime the system does anything it's flooded with interrupts, top shows a constant 20-40% interrupt time. I'll try installing on a hard drive and hopefully that will improve performance.

Does your Sun/2 have the ethernet adapter that's mentioned in the FAQ as only able to receive at most two packets before dumping them on the floor? I imagine that would make it especially painful.

Broadly speaking it's my recollection from 20 years ago that I used to actually see better performance running off of NFS than local disks when fooling around with the Sun boxes, but these were SparcStation 5-level machines, not Sun2s.

 

[resman]

NetBSD 9.2 and the long boot time was mostly from generating the fontconfig cache which was about 6 hours. Even after booting the system is barely usable because of makemandb running in the background for several hours. I think a large part of the slowness is due to running everything over nfs so anytime the system does anything it's flooded with interrupts, top shows a constant 20-40% interrupt time. I'll try installing on a hard drive and hopefully that will improve performance.

I submitted a patch something like 15 years ago to the NetBSD 68k kernel to back off the frequency of the timer interrupts. It was something like 100 Hz and my poor Mac SE/30 at 16 MHz couldn't get out of it's own way. Obviously it wasn't accepted.

 

[Eudimorphodon]

I submitted a patch something like 15 years ago to the NetBSD 68k kernel to back off the frequency of the timer interrupts. It was something like 100 Hz and my poor Mac SE/30 at 16 MHz couldn't get out of it's own way. Obviously it wasn't accepted.

I guess FreeBSD implemented a tickless kernel in 9.x, looks like it's been at least thrown out as an idea for NetBSD, but maybe not anything more than that. Although I guess I don't know if the mac86k architecture has appropriate programmable timer hardware to make that work in the first place.

 

[tokenalt]

You can add makemandb=NO and fccache=NO in /etc/rc.conf before first boot.

Thanks for this. Disabling those and ldconfig has greatly improved boot times.

Does your Sun/2 have the ethernet adapter that's mentioned in the FAQ as only able to receive at most two packets before dumping them on the floor? I imagine that would make it especially painful.

Broadly speaking it's my recollection from 20 years ago that I used to actually see better performance running off of NFS than local disks when fooling around with the Sun boxes, but these were SparcStation 5-level machines, not Sun2s.

Sorry if it wasn't clear before but I'm using a 3/80 not a 2/50. It has the same lance ethernet chip the sparcstations have altho I doubt it's getting anywhere near the same performance as the ss5.

I ran ubench on the machine and got a blazing fast score of 107 cpu, 91 mem, 99 avg. For comparison an ss10 w/ sm30 gets 1105 cpu, 2554 mem, 1829 avg. It absolutely blows my mind the difference in performance of these two machines that were released just 3 years apart.