FreeBSD Support

[FalloutZone]

Support FreeBSD Binaries and System Calls

Expected Result:
Run FreeBSD on the subsystem

FreeBSD is a very popular server operating system with deep roots in web and cloud development. Please support FreeBSD in the Windows 10 "WSL" feature.

[Liam Crabbe (drsn0w)]

This is Windows Subsystem for Linux, not FreeBSD or other *NIX environments.

[FalloutZone]

Yes, you are correct - WSL is for ELF/Linux. This is a request to make this, or a similar system on Windows 10 WSL to support other open source *NIX environments - specifically FreeBSD.

[[COMMAND.COM ... .. .] (kilitary)]

they even can make frozen versions of lib and loading elf or a.out will be just a user ring asm code.
but in reality i do not clearly know how WSL is implemented because i do not need this. wsl is good.

[Sunil Muthuswamy (sunilmut)]

FalloutZone - Thanks for the feedback. What you are essentially asking for is "WSB" (Windows Subsystem for BSD). User voice ticket would be a great way to provide that feedback. Something that we look at for our future prioritization.

[matt wilkie (maphew)]

"This UserVoice instance is no longer available."

[eeboaz]

I would like to see this as well. I prefer the FreeBSD operating system over Linux for server and development use cases and this would make Windows 10 a much more viable option for a development platform for me. I will look at the User voice ticket link provided.

[Federico (iz0eyj)]

it may be that it becomes useful in a possible version for use on servers, but the use of *BSD in desktop machines is really very limited.

[Onepamopa]

I would also like to see Windows Subsystem for FreeBSD.

[eeboaz]

it may be that it becomes useful in a possible version for use on servers, but the use of *BSD in desktop machines is really very limited.

There's no good reason to arbitrarily permit *BSD only on servers but not on desktops.
The use case for WSL is not the server market - it's the developer desktop market.

BSD is much the same. This would let me have one system at work where I have to use Windows for corporate stuff (exchange/email/calendar, visio, etc.) and BSD where I code for servers running FreeBSD
Apple is continuing to make a mockery of macOS and isn't really viable any more.

[Federico (iz0eyj)]

eeboaz:
I have absolutely nothing against BSD, but his primary field of use are the servers and network devices. The WSL we have now is not suitable for use on any of the two types of object, and the effort needed to create in "WSB", at least in my opinion, it would be totally unjustified.
If later the characteristics of WSL will change in a direction that will be suitable for those uses, then it will change my opinion.

[eeboaz]

Federico (@iz0eyj)
Why are you insisting on this view point that we're wanting to run production systems on WSL/WSB?

This is exactly counter to the purposes of WSL in the first place, straight from the FAQ:

Who is this for?

This is primarily a tool for developers -- especially web developers and those who work on or with open source projects. This allows those who want/need to use Bash, common Linux tools (sed, awk, etc.) and many Linux-first tools (Ruby, Python, etc.) to use their toolchain on Windows.

[Federico (iz0eyj)]

eeboaz:
where you see reported the BSD say in what you brought? if anything at all WSL documents contain references to Linux and exclusively to Linux. For a project to be successful it needs there are users who then use it ... the development team has a cost, are not eople who work for the glory.
What do you prove by that passage? which it is directed ONLY to the developers? is not the case, this is what he would like some of them but it does not. For any developer who receive WSL exist thousand and thousand ordinary users; It will be sufficient for an ordinary user of a hundred active WSL to see them dominate much the number of developers.
There are no end-user Windows interested in BSD, they want Ubuntu Linux.
WSL is not a product for the community "free software" but it is a Windows component, whose main target will be Windows users, be they end users or developers.

[eeboaz]

Federico (@iz0eyj)

There are no end-user Windows interested in BSD, they want Ubuntu Linux.

I did't know you spoke for the entire Windows community. Right, you don't.
You're not going to dissuade me from being vocal that I would like this support.

We're just asking the development team at Microsoft to consider supporting FreeBSD in addition to Linux when they've reached their primary goal of supporting Ubuntu Linux. I don't see how having options harms you at all, so please be considerate and let me and others ask for the same option: using the Windows subsystem to support the user-land of the operating system we prefer to use for our development work.

[rustyx]

I'm surprised this "issue" is still open - this project is about Linux emulation, after all.
BSD user mode is way more tightly coupled to the kernel than Linux. Given that, plus the dynamic nature of the FreeBSD ecosystem, emulating a BSD kernel would be an exercise in futility.
Which is no problem at all because there is very, very few reasons to emulate BSD anyway.
You can never use it for testing - if your code doesn't work, you'd always want to try it on a real BSD.

It's not about having the option, it's about spending hundreds or thousands of development hours to create and maintain something that is marginally useful.

[Adam Seering (aseering)]

G'day all -- there are some valuable opinions here, but could I request that we keep this thread on topic?

The topic here is a request for support for a BSD userspace. That is a valid technical request. It is up to the WSL team to pick it up or not. I, for one, trust them to have sound judgment. If you have supporting use cases or technical commentary on making that happen, please post it here.

If, on the other hand, you think this is a totally silly idea, stop replying and bumping up the ticket's post-count :-) According to this repository's README, issues remain open until the "discussion has run its course." So I encourage you to resist the urge to correct the thing that you see as Wrong On The Internet, and let the thread come to a stop.

[therealkenc]

User Voice is here.

[ruffianeo]

The one reason I would love to see FreeBSD support (even instead of linux) is that I have a deadly GPL allergy and I try to stay clear of such (legal) issues by avoiding Linux whenever I can. The "mostly for network and servers" argument does not cut it IMHO. FreeBSD is THE Unix for people who hate GPL.

[Deleted user (ghost)]

Same

[Deleted user (ghost)]

[[COMMAND.COM ... .. .] (kilitary)]

like a char m

[Deleted user (ghost)]

knock knock

[[COMMAND.COM ... .. .] (kilitary)]

обьект обнаружен ... обработка ... но нам нахуй он не нужен

[Nikolai Zamega (nikolaizamega)]

How to patch KDE2 for FreeBSD? 😂

[Rishikeshan Sulochana/Lavakumar (Work) (ris-work)]

I'd love to see any BSD support as well. NetBSD already runs as a RUMPkernel (and has been compiled to JS as well!) so I suppose it should not be very different to what's done for Linux. Moreover, I suppose the license is far more lenient, too.
Since it already uses a hypervisor anyway, it should at least be bootable?

[zhanghj (netjune)]

I think it should be easy to support FreeBSD using hyper-V like WSL2.

[Trung Nguyen (trungnt2910)]

For the NT kernel driver devs out there:

As of Windows 11 23H2, it is possible to create custom Pico providers/library OSes on Windows.

Currently I have created a toy subsystem that can print "Hello World" using a custom syscall ABI. In the mid-term future, I might attempt OSes of mid-complexity like Haiku OS, before proceeding to newer Linux/Darwin/FreeBSD in the long run.

I'm current wading through the NT kernel internals and learning new things every day, join me if you can contribute or just wanna have a chat.

[Deleted user (ghost)]

[Nolram (Nolram12345)]

+1 on this. BSD for WSL (WSU?) would be awesome.

[Xzenor (xzenor)]

+1 on this. Not that I expect it will ever happen but still... Worth a try

[DEVIN (DeviNoles)]

+1

[pleasesendmeloadsofvegetables]

Can we get an official stance on this? The last time a maintainer posted in this discussion was in 2017. I would love to use WSBSD (or whatever it would be called) instead of WSL.

[Fuwakami Mana (ManaHime)]

With wsl becoming mostly open source, I wonder if this means that wsbsd might become possible? ＾＾

[WesleyKhoiVo (VOM7HC)]

Thank Trung Nguyen (@trungnt2910), I also think WSL 1 is better performance than WSL 2 (which based on Hyper-V), but one will stuck that Windows is never be POSIX compliant so if you have app or program call POSIX syscall API like fork() this will meet unexpected behavior. How your handle it? I still see micro kernel concept like WSL 2 is suitable for POSIX compliant if run binary without modify

I saw all your comment and posts related with "Revive Astoria", LXMonika and other comment relate WSL discussion here. I am also interest all of these. Because I very interested in Windows Programming. But I have different view of this path, like WSX (Windows Subsystem for XNU) which will run some app like xcode like darling, which will contain some code from WinObjC and ipasim.

Are you Vietnamese? Can we share contact together?

[Trung Nguyen (trungnt2910)]

WesleyKhoiVo (@VOM7HC)

but one will stuck that Windows is never be POSIX compliant

Windows used to be POSIX-compliant with its own POSIX subsystem. POSIX itself is not a big issue, since it is a rather high-level standard, leaving a lot of room for implementation-specific behavior.

In fact, WSL1 handles plain POSIX syscalls quite well, since they're well maintained in WSL1's prime in 2016 - 2017. It is only when Linux-specific syscalls/features (e.g. epoll, /proc files) get involved will things fall apart.

How your handle it? I still see micro kernel concept like WSL 2 is suitable for POSIX compliant if run binary without modify

WSL2 is not a micro-kernel in its pure sense. It is not a component dependent on Windows, it is an entirely separate virtual machine (just like VirtualBox) with some magic sockets.

I do agree that to get full Linux compatibility (not POSIX), we need to involve the full Linux kernel. This is what I plan for lxmonika. The approach might be difficult for Windows due to licensing (building a lxcore.sys based on Linux will get winload.efi infected by the GPL) and other issues, but is possible for an open-source project with enough time investments.

But I have different view of this path, like WSX (Windows Subsystem for XNU) which will run some app like xcode like darling, which will contain some code from WinObjC and ipasim.

In my opinion, library-level compatibility like what Darling and ipasim are doing will not be sustainable. Interfaces provided by user-mode libraries are seemingly infinite, they are much wider than that provided by the kernel (just a few hundred system calls and maybe a hundred more magic ioctls. My vision of the Windows Subsystem for XNU would be something that uses the full XNU kernel as a Pico/lxmonika provider, then tries to run unmodified macOS libraries and frameworks.

Are you Vietnamese? Can we share contact together?

If you're interested in contacting me, try joining the Project Reality Discord. We have a dedicated Vietnamese channel.

[WesleyKhoiVo (VOM7HC)]

Thank Trung, sorry make you mis understand,

In my opinion, library-level compatibility like what Darling and ipasim are doing will not be sustainable. Interfaces provided by user-mode libraries are seemingly infinite, they are much wider than that provided by the kernel (just a few hundred system calls and maybe a hundred more magic ioctls. My vision of the Windows Subsystem for XNU would be something that uses the full XNU kernel as a Pico/lxmonika provider, then tries to run unmodified macOS libraries and frameworks.

WinObjC is port for ObjC runtime, mostly still for user land mode
for kernel mode, still base BSD and OSFMK

but that mean I show to you my view my idea

[CielRuby (kevincrans)]

I was for fun making up something similar recently in the ravynOS community: running a macOS/Darwin BSD userland on top of the ReactOS NT kernel.
So maybe RSBSD is an idea, in the hands of the community.
I agree and think that WSL2 is not WSL, and WSL1 could have been faster if done right, & more compatible with a lot of effort.

[Trung Nguyen (trungnt2910)]

I think we already have most of the puzzle pieces for a WSL1-like approach for FreeBSD on NT 10.0:

1. A (fairly) stable way to create non-WSL Pico processes: https://github.com/trungnt2910/lxmonika
2. A way to statically embed an ELF shared object into a PE executable: https://github.com/trungnt2910/REPEAT
3. A way to produce a library containing the high-level layers of the target kernel with no hardware-specific dependencies.

• For Linux, we can already harvest the objects from a UML compilation.
• I am not sure if something similar is available for FreeBSD?

All it takes now is $200 for ChatGPT to write some glue code and an engineer motivated enough to craft the prompts.

Alternatively, I have heard of someone writing a fully-userland compatibility layer for OpenBSD, which proves that you may not need kernel intervention to emulate a Sys-V ABI on Windows.

[Coolis1362 (Coolis1362)]

Don't Worry I Will Make WSBSD, Starting With FreeBSD Support At WSBSD

[WesleyKhoiVo (VOM7HC)]

Sorry coolis, I reviewed your code not do anything here. With only managed code c# it is not subsystem.

[Trung Nguyen (trungnt2910)]

🤫 Don't break the kid's fantasy... He's doing quite well for a 9-year-old.

Everyone knows that as of now, the only known creators of Windows subsystems are Microsoft and Project Reality 👀

[WesleyKhoiVo (VOM7HC)]

Haha Trung, I don't extinguish the fire inside anyone, but some one just have background python script and their backend code is managed code with c# with support LLM and vibe coding style could not produce any real product here.

Microsoft has very long term with they interix and now with WSL, if one guy can make subsystem better them, he should pointed as CEO M$ instead monkey coder here 😁

[Trung Nguyen (trungnt2910)]

managed code with c#

This is not an issue. NativeAOT is quite powerful and can produce valid PE files. With some support scripts, they can be signed and converted to a driver.

with support LLM and vibe coding style

This is the real issue. Regardless of what other people say, AI is too stupid to do any real work, especially something that requires accuracy and strict compliance like implementing a subsystem.

if one guy can make subsystem better them

I bet many people can. The average PhD candidate in Computer Science can probably code better than the current M$ CEO.
But M$ is an enterprise, not an academic institution. They are all about money and influence, not about advancing technology.

[WesleyKhoiVo (VOM7HC)]

This is not an issue. NativeAOT is quite powerful and can produce valid PE files. With some support scripts, they can be signed and converted to a driver.

Haha, yeah you can produce nativeAOT or use unsafe classic c# or deprecated .net native to produce native code. But I just "provoked" him/her/them code because it only console print

I bet many people can. The average PhD candidate in Computer Science can probably code better than the current M$ CEO.

I know dude, some one can work. But how they can have money from that. This is business choice. I just "provoked" him/her/them when him/her/them release WSBSD just 2 week after M$ open source it.

[pleasesendmeloadsofvegetables]

Once again, I'm asking for an official stance on this. As I mentioned previously, the last time a maintainer posted in this discussion was back in 2017. Especially now that WSL has been open-sourced, an official stance would be great.

[Pedro Giffuni (pgiffuni)]

The WSL2 approach is being worked on (outside Microsoft):
https://github.com/BalajeS/WSL-For-FreeBSD

[Maseru898]

Seems like Development halted basically after you posted this :D

[Fuwakami Mana (ManaHime)]

There seem to have been a talk at BSDCan?

[Maseru898]

Any Update? Been 9 years now, i think enough time to make it work