OS/2, eCS & ArcaOS - General > Events
Warpstock 2025 USA - The end of my journey with ArcaOS
Roderick Klein:
First of all I listened to many hours of the Warpstock stream and I do not think that Lewis said it such a way as one line answer. Like ooh who needs Wifi, in my opinion Sigurd did not picked up the message 100% correctly. David mentioned if he could hand over to another developer he would welcome that. He did not state (from what I heard) he would stop development.
And while I perfectly understand everybody is entitled to his opinion and call stuff excuses. They way I understand the idea is Arca Noae works on drivers to keep ArcaOS to boot on new hardware. And that is logic order in my opinion. So as David and Lewis mentioned. IBMKBD.SYS was updated to support systems without a 8042 controller, we take NVME support for granted. USB 3.0 took waaaay longer then every expected. And now David is working an update to ACPI as discussed by Dmitry. And why do we not have Wifi drivers.
We you can port drivers to another OS in two ways.
1. You rewrite the code to match OS/2. from Linux/FreeBSD that is very, very time consuming.
2. The second option is that you write a driver library so you work more efficiently.
Option two we started with when David still worked on eCS at Mensys and worked on the Multimac driver system. With library he could with almost changing no code and push out Linux kernel NIC support on OS/2 within a few hours per driver. Work was resumed to push this towards Linux and when it was not far from being finished. Well the Linux guys changest the kernel in such a way that broke the Multimac library to a high degree. And that was the end of the dream of portable wifi drivers, blown. I once heard from David at one of the Warpstocks
then one binary for the Wifi Intel driver (that not working yet, it was not finished). One binary was 400 KB in size. I think that is just the size Uniaud32.sys.
Wifi drivers are so waaaay, waaaay much more complicated then ethernet drivers.
So after the Linux option went away David started from scratch on Freebsd. Some concept but it takes time to build such library. Every call the driver source make need to mapped somehow to work on OS/2. So ethernet drivers started to appear. And when stuff got closer for Wifi with a FreeBSD library well the FreeBSD modified the kernel sources.
Now David his work to port the 2.5 gigabit drivers is slowed down also as he encounters option 1 I mentioned. Porting code. All of that is way more labor intensive.
So for people to sit there that nothing was done on wifi drivers etc.
Cutting long story short. Serious efforts where made for wifi drivers and mostly failed because of bad luck. If Linux or FreeBSD did not modify its kernel sources we would have had the wifi drivers. Indeed wifi drivers are one of the most complex driver types around just about. Panorama, AHCI, NVME, ACPI and USB is so much easier, but why is that ? Well the interfaces are to high degree standard.
Now porting the new Intel and Realtek ethernet drivers might also take longer because of the same changes to FreeBSD kernel.
Its not recompiling the code....
Just my humble opinion.
Roderick
Roderick Klein:
--- Quote from: Martin Iturbide on November 15, 2025, 06:29:48 pm ---Hello
--- Quote from: andreas on November 12, 2025, 04:01:10 pm ---wished i was a programmer to contribute to the development of a WIFI-driver.
What about the idea to start a project to develop one together - online with the help of all who have the knowledge, time and interest - step by step. Would that be feasible?
--- End quote ---
Even that the Andi B's comments are right, I was wondering if a more dirty/non-elengant/hack aproach can be worked out to have wifi drivers with less effort. Like just focusing on some basic wifi first and maybe not fully functional first. Maybe going back the 32bits windriver wrapper idea, or focusing on some USB wifi dongles.
Regards
--- End quote ---
Read my lengthy reply to Sigurd if it was all that simple, really I wish it was. Sadly its not. I have been looking at all of this.
What we need understand as community is what I described below:
You effectively have two approaches.
1. Port a driver, copy paste chunks and get a driver.
2. Come up with library or wrapper that does this. This what David tried twice with Linux and FreeBSD. But it backfired because simply a case of badluck. The bad luck that Linux changest the kernel somehow and then it happened again with FreeBSD.
The win32 wrapper is complete dead end. Simply because no more 32 bit drivers are being made for Windows. You know when Willebald Meyer created Genmac it was a lot of work I am not discounting that at all. But to some degree he also got lucky and here is why. The NDIS network stack on OS/2 and Windows where pretty close (no surprise of course).
One of the changes is that NDIS stack on Windows started to change that much (and the driver as well). To an extend keeping Genmac up with the changes on Windows became to much for one person to keep up with.
I wonder if people even fully grasp the parallels between Multimac and Genmac. To put it differently Willibald Meyer looked a generic solution to not write the drivers.
But wrote drivers to support the Windows drivers. My conclusion Genmac stopped working simply because the technical changes between Windows and OS/2 become to big.
ReactOS I think also only supports 32 bit wifi drivers, not 64 bit.
ReactOS seems be in the same boot as we are.
https://reactos.org/wiki/Supported_Hardware/Wireless_Network_cards
Roderick
Roderick Klein:
--- Quote from: Andi B. on November 12, 2025, 04:15:01 pm ---
--- Quote ---who have the knowledge, time and interest
--- End quote ---
* who have the knowledge
* time and
* interest
Tell me when you have a candidate for only one of the above 3 questions.
Seriously, I've been the maintainer of xwlan and our port of wpa_supplicant for about a decade. Although I've fixed and extended a lot and invested countless hours, even these two project are far from what we would need today. And that's only an interface to a driver. I know porting or writing a wifi driver would be a lot of more (huge more) work. From my experience the majority of 'non programmers' don't even get feedback. Not to talk about helping out with testing or writing readme's..... So dream on.
--- End quote ---
Look I am not an expert but heck you ask Copilot.com at least as *some* form of start. This list comes pretty close to what is needed to write an OS/2 wifi driver. Maybe not all is needed. But let this list sink in:
1. Device Identification & Initialization
PCI/USB/SDIO bus interface code to detect the chipset.
Proper vendor and device IDs so the OS can bind the driver.
Initialization routines to set up registers and memory mapping
2. Firmware Loader
Many WiFi chipsets require proprietary firmware blobs.
The driver must include routines to load firmware into the chipset at startup
3. MAC Layer (Medium Access Control)
Implements 802.11 protocol logic (association, authentication, encryption).
Handles packet framing, retransmissions, and QoS.
4. PHY Layer (Physical Layer) Control
Manages radio tuning, modulation, and channel selection.
Provides APIs for setting frequency bands and transmit power.
5. Network Stack Integration
Hooks into the OS networking subsystem (e.g., Linux cfg80211 and mac80211 frameworks).
Provides standard interfaces for scanning, connecting, and data transfe
6. Configuration & Management Utilities
User-space tools or APIs (like iwconfig, NetworkManager, or custom utilities).
Support for WPA/WPA2/WPA3 security via integration with supplicant software
7. Interrupt & DMA Handling
Efficient handling of hardware interrupts for packet transmission/reception.
DMA (Direct Memory Access) setup for high-speed data transfer.
Skipping power management
9. Error Handling & Logging
Debugging hooks, kernel logs, and recovery routines for failed transmissions or firmware crashes.
Doug Clark:
A few thoughts, FWIW, about the wifi discussion:
1) I was sitting next to a guy (Grant) at Warpstock who was using his iPad to provide internet to his Lenovo laptop running AOS 5.1.1 (I think it was through his iPad but might have been through his iPhone). So it is possible. If anyone wants the details I will email him and ask - if he isn't already reading this topic.
2) Grant is an assembly language programmer - his preferred language. One of his goals is to get bluetooth running on OS/2. He has an ODB2 car scanner that he wants to be able to read using OS/2. So we MAY get some bluetooth support, for at least some devices.
3) I also watched multiple people messing with USB to Ethernet adapters to get internet on their laptops: some were successful, some were not. So there ARE some that work. Like almost all devices, you have to be careful when you purchase to get something that works with the drivers already available for OS/2.
4) As Dave Leo pointed out, there currently are limited resources available that can be applied to OS/2 related problems, like drivers and a web browser. And some of those previously available resources are now less available because of world events. So someone has to choose which problems get which resources.
5) There are devices which can be purchased today which enable laptops running OS/2 on bare metal to access the internet. Those solutions are not as convenient as turning on the radio switch on the laptop and having internet access. But they do exist, they do work, and they are relatively affordable.
6) I think that is what Lewis was trying to express at Warpstock.
So yeah wifi isn't here yet, and it very well may not be here for some time to come. But from my perspective I think that the decisions on where to apply resources have been pretty spot on. AOS can run bare metel on new machines - which is a pretty amazing thing.
Roderick Klein:
--- Quote from: Doug Clark on November 20, 2025, 12:12:07 am ---A few thoughts, FWIW, about the wifi discussion:
1) I was sitting next to a guy (Grant) at Warpstock who was using his iPad to provide internet to his Lenovo laptop running AOS 5.1.1 (I think it was through his iPad but might have been through his iPhone). So it is possible. If anyone wants the details I will email him and ask - if he isn't already reading this topic.
2) Grant is an assembly language programmer - his preferred language. One of his goals is to get bluetooth running on OS/2. He has an ODB2 car scanner that he wants to be able to read using OS/2. So we MAY get some bluetooth support, for at least some devices.
--- End quote ---
Anybody gots Grant contact information ? If so would somebody please send him my contact data (or he contact me) on roderickklein at xs4all.nl.
As for Bluetooth support. I am no expert in this but this might be a bit a big one to get working as I get the impression you would need a Bluettooth to support a bluettooth device. You need to for exaple talk to Bluetooth transmitter in your PC. So I wonder how much work it would be.
Roderick Klein
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version