ltning
@ltning Wow, 128MB of RAM on a 486. I bet that's a huge part of how it's doing as much as it's doing. 😆
Yeah. Having RAM helps. But only half of that RAM is even cached ;) I should find a similar board that has 1MB cache.
Worth mentioning I have NetBSD running on 16MB as well, but I'm not going to run any kind of server on that. ;)
@ltning
What do you use it for?
Mhhh.. #snac crashes when it start's in my Dual Pentium II system. It's performing at start a heavy job "started deferred data integrity check", which does lot IO.
When I turn off the traffic from outside, it seems to run (takes ages). But when I turn on the traffic to it, it crashes with
`Illegal instruction (core dumped)`
I took the installation over from the PIII system and activated the MP #OpenBSD kernel. (Do you really have to cp it from the CD by yourself??)
@DeltaLima
Recent versions of OpenBSD automatically install the right kernel. Are you running a fairly recent version?
@FritzAdalis 7.8 downloaded the iso two weeks ago.
I took the SSD over from an P III single core system, where I installed it. I thought OpenBSD should be capable as well to get swapped around similar hardware.
@DeltaLima
Usually it can. I'm not sure of the technical reasons for having single vs. multi processor kernels any more, but they should at least boot.
(I guess I read what you wrote backwards from what you meant.)
@FritzAdalis I've tried the "normal" bsd kernel without mp support again, and it crashes there too :/
Looks like I have to do some more debugging.
@DeltaLima
Yeah, it shouldn't crash. Maybe try ktrace/kdump on it and see where it dies, or gdb on the core file if you know how to use it. (I do not.)
@FritzAdalis I tried ktrace+kdump but didnt get something out of it. At the end, it gets a SIGILL, after trying to read a json file. I then did a dump with tracing all childs and saw it seemed to read a lot of certificates - so I assume it's doing some SSL stuff and fails there (?) 🤷
```
19370 snac PSIG SIGILL SIG_DFL code=ILL_PRVOPC addr=0xa0284f0 trapno=0
19370 snac STRU struct stat { dev=10, ino=2728365, mode=-rw-r----- , nlink=2, uid=1001<"snac">, gid=0<"wheel">, rdev=109...
```
@DeltaLima
Ssl... I wonder if it's expecting the cpu to have an instruction that didn't exist then.
@FritzAdalis Idk - @ltning is running two servers on similar machines, one with AMD 5x86 and one with dual Pentium on NetBSD. He's also running a patched version, but from what I understood (C programming noob) there was nothing directly ssl related in those commits.
I'll try this fork next just for fun.
@DeltaLima @ltning
Hm, Pentium and 5x86 wouldn't have any instructions that a PII doesn't. Could be something that was ripped out of libressl but netbsd is I think still on openssl. Or I'm completely off base and it's just a regular bug.
@ltning @DeltaLima
OpenBSD source will always output the same code independent of the current cpu capabilities, but other code may not.
@FritzAdalis @ltning I think I'll try NetBSD next as troubleshooting step, hope I will find time at the weekend for it. Thank you so far! :)
You may want to make sure you disable encrypted swap unless you have a ton of RAM. Someone seems to think that if you're on x86 you always have CPU to spare for encryption :D
Also, pkgin is fantastic and many (more than on OpenBSD) packages are built so they work on even an i486.
Also amazing that this thing did not OOM-kill processes all over the place; even before I started compiling I did a pkgin upgrade on the poor box, which took .. a few hours. I'm all outta swap, real low on RAM, and yet it somehow got through this .. #NetBSD for the win, I guess.
Says top:
Memory: 56M Act, 27M Inact, 11M Wired, 15M Exec, 820K File, 2076K Free
Swap: 128M Total, 128M Used, 4K Free / Pools: 23M Used / Network: 23K In, 34K Out
History