Hello, I am still wrestling with GHC 9.8.2 via pkgin on OpenIndiana (illumos). I figured out what the issue was with the C++ lib, and temporarily copied libgcc_s.so.1 into the directory where pkgin expects it. gcc-13 installation succeeded, and GHC installed with 0 errors and 0 warnings, albeit broken. Upon running ghc, an error message is

immediately returned which reads, "ghc-9.8.2: internal error: timerfd_create: No such file or directory (GHC version 9.8.2 for x86_64_unknown_solaris2)". I am struggling to make sense of this

timerfd.h in in /usr/include/sys/

josh's comments here might be a helpful lead, if you'd missed it earlier: log.omnios.org/illumos/2024-12-26#1735239330-910753

Just reviewed the log. I did miss that. Thank you

hiya i am brand new to illumos, how do i make network configuration persistent?

what distro are you using and how did you configure it initially?

i'm using helios, which had instructions to run `ipadm create-addr -T dhcp igb0/dhcp`

if /etc is persistent on it (I'm not sure offhand, but others should be able to answer then when they're around), then using dhcp should be persistent

or did you want to manually configure an address?

no i am ok with dhcp, but it didn't pick up an address after i rebooted

it seems like /etc is persistent if i'm reading the output of `mount` correctly...

snek: How did you install it? FWIW, we recommend people use OmniOS (which is quite similar) for general purpose stuff. The installable Helios media mostly exists to support engineering efforts on the Oxide rack software.

i followed the physical install instructions in the engvm repo

i'm just trying to see how far i can get with running the various oxide rack components :)

Fair enough!

When you ran "ipadm create-addr" was when you were booted from the install media

Or after you rebooted from the installed disk and logged in etc

it was after

though i just rebooted again and it seems to be working now

Great!

so probably pebkac of some form

It's possible, but I wouldn't assume. Software can always have bugs

i ran `pkg install cmake` and it claims to have installed it, but its not in my path... is this expected?

hm did i just need to add `/opt/ooce/bin` to my path? doing that for now until someone here tells me its a mistake

I would not be surprised if you had to do so manually there.

How would one go about porting a Linux or BSD NIC driver to Illumos? It is still nagging me inside that there are drivers for this newer NIC and I want to figure out how to get it working under Illumos but not well versed in driver development.

It's definitely harder than trying to compile the drivers and changing a few things around but if it is something I could do with some training then I would like to try it.

the driver interfaces are different enough that you're probably better off writing a driver from scratch using an existing illumos driver as a template, at least in the upper layers of the driver. Avoid copying code from linux but it's ok to look at a linux driver for hints on how the hardware works.

There are a few ways to approach it depending on your background.

The first thing I'd do is understand what the GLDv3 interfaces are and spend some time reading the drivers to understand the distinction between what's required to configure and use the device, what's being done for different entry points, what's DMA, etc.

I would take a look at the mac(9E) manual and the updated Intro(9E) manual as well (which is a bit behind online for a reason we need to track down)

i always wished drivers were written in a more platform-independent way when i was working on my os 😔

There are lots of trade offs to what I call the "common-code" approaches.

It ultimately means parts of code look like someone's OS and sometimes you aren't given the abstractions you need.

Even if they attempt to make it more library-ish.

Fair play. Thank you for the guidance. I will start with the manuals and you both make good points about using code from another OS for more than hints. Appreciate the help.

Otherwise the other appraoch is to first learn how to build the kernel on illumos and make some unrelated change.

So you can get some muscle memory built up.

That may be a better starting point for me and was my initial thought. How does one compile new modules for Illumos in the first place?

So I would start with illumos.org/docs/developers/build.

And probably make a change to something that exists.

And prove that your change has happened on a target system (whether hardware, a VM, or something else).

Once you have something like that working, the incremental bit towards adding something new won't be so bad.

Nice! Thanks again for the help. It would be exciting to have a bunch of these minipc's running OmniOS or OpenIndiana for a...what's a word for a mini-cloud? Vapor seems like vaporware, mist perhaps?