this post was submitted on 09 Mar 2024
399 points (88.3% liked)
Gaming
3127 readers
827 users here now
!gaming is a community for gaming noobs through gaming aficionados. Unlike !games, we don’t take ourselves quite as serious. Shitposts and memes are welcome.
Our Rules:
1. Keep it civil.
Attack the argument, not the person. No racism/sexism/bigotry. Good faith argumentation only.
2. No sexism, racism, homophobia, transphobia or any other flavor of bigotry.
I should not need to explain this one.
3. No bots, spam or self-promotion.
Only approved bots, which follow the guidelines for bots set by the instance, are allowed.
4. Try not to repost anything posted within the past month.
Beyond that, go for it. Not everyone is on every site all the time.
Logo uses joystick by liftarn
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I don't know where the hell you got that information but that's not how proton works. There is no "fake kernel" it's not a virtual machine or an emulator it's just a translation layer that translates Windows syscalls into linux syscalls
The syscall translations that would go to the nt kernel, can be seen as a "fake kernel", no?
Wine has a process that works as a substitute for the Windows NT Kernel. How that works in detail, which calls are abstracted with an internal model, and which are mapped on to Linux kernel calls, is a bit silly to get hung up on, no?
I think it's perfectly fine to call that concept a "fake kernel". I don't know what you'd need in order to qualify more?
"Just"?. No. It also has an internal model. Which system calls end up as Linux syscalls, and how, is not a stateless translation. The NT kernel is modeled. And although you are right in your straw man argument that it isn't a "virtual machine", or an "emulator". Neither of those are a requirement for the concept of a "fake kernel" either. Seems a bit rude to go so balls out hard against it, as you did.
None of the things needed for a true kernel level anti cheat are in a translation layer, some of them just can't be. It's why games with actual kernel level anti cheat have never worked in proton and never will. Historically the only way to play those games is either be on windows or use a vfio virtual machine (which also probably won't work even with tons of vm hiding techniques depends on just how sensitive the ac is). Wine/proton simply can't translate the upper parts of the nt kernel needed for it.
Some games that use EAC, BattlEye and GameGuard, work fine in proton. Afaik, whatever these do and are abstracted to, or is offhanded to some linux native process, it's still all running in userspace. I'm sure this relies on individual game developers playing along with it, and not 100% "proton emulating the nt kernel" in order to "fool them". Is this the point you're arguing? That it won't be possible by a purely proton/wine translation layer?
If you know details on how exactly this works, or want to point to some resource on this, I'd be happy to read more about it.
My guess is that there is nothing technically impossible about fooling a rootkit by faking whatever syscalls from the game, but that it becomes a impossible task to maintain, as the AC developers can make minor changes that would require a lot of hard work to "emulate". I'd love to learn more, but it was hard to find good resources on this.
EAC has an explicit linux support that valve worked on them with, it's not full kernel AC. same with battleye and GG those are not full kernel root anti cheat implementations. I can't point at any specific documents unfortunately but the wine/proton irc channels are public and lurking let's you learn a lot as they talk through issues with games and anticheat.
In order for linux to support kernel level AC a module for the Linux kernel would be needed. And i doubt Linus will ever allow that lol
This is... correct. That in order to support kernel level anti-cheat on Linux, you need a kernel module. But that statement is a tautology.
An NT kernel AC running through Wine, and whether or not it "works" doesn't predicate on a Linux kernel module being loaded. All it needs is the correct handling of whatever the NT kernel would communicate to the running game, and handle whatever that callback is through some other mechanism that passes the checks.
Most AC software have Linux native clients, and that's what this "some other mechanism". And whatever that is in practice, should, with enough reverse engineering, be technically possible for proton/wine to do as well. It's all running on userland after all. I assume that this is not an easy task to do at all, which is why the only realistic approach is for AC developers to actually be on board, and instead just compromise on the weaker level of anti-cheat protection, compared to what you'd get with a kernel module. As far as I understand, this is the case for GG, BattlEye and EAC. Not all games work, because it depends on the developers "allowing it".
And as for what the future might bring. I expect that with Linux gaming becoming more popular, it's only a matter of time before a Linux AC is implemented as a kernel module. Also, neither Linus, nor anyone, need to whitelist a kernel module for it to be loaded. The only one that has an ultimate say there is you, the user.
I mean sure they could make it a dkms module and have the user install it along with the headers but it's never going to be out of the box supported on linux was what I meant by that
As for the rest, there is a limit to what can be emulated within user space. There are are certain calls in NT ring -1 that would require similar privilege on the Linux side to translate which i doubt wine would ever do for a vast multitude of reasons
Why would that be the case? I have to look this up and read more about it, because I don't see why that needs to be the case. I'm also not sure if this is still in the context of AC "rootkits". Because if so, I imagine the security model goes something like this
AC RootKit: Can observe app processes and all memory usage, and modify anything at any time. It observers processes for known cheats, and reports this to the game, either with a callback the game registers, or by directly modifying the game memory.
Wine: Runs in userland. Syscalls are "intercepted" as with all other windows API calls. The NT kernel doesn't exist here. Wine just tries to answer those calls as if it did.
Game executable: Has some mechanism to talk to-from the rootkit. Likely that the RK itself, since it monitors processes, hooks straight into the game exectuable by either manipulating the memory required for the game to say "ait, RK said you're cool", or something like that.
Game executable running in Wine: Runs in userland, and wine has already talked to the Linux kernel and allocated memory. To the loaded game executable running through wine, the memory can be manipulated the same as a rootkit could, because after all, the wine process is the parent process of that memory range.
So, what mechanism is it that an AC RK does, that, from the perspective of a user process running on Wine, cannot be done unless actually coming from the Linux kernel? I honestly cannot think of anything.
Or rather... only possible way I can think of is a "cryptographic guarantee", in some secureboot based signature and communicating with a remote service in order to authenticate the RK , which the game executable also confirms. Something like that. But this isn't the case for any of the AC RKs afaik
Yes, and part of these make the anti cheat believe it's running in a kernal.
None of the things needed for a true kernel level anti cheat are in a translation layer, some of them just can't be. It's why games with actual kernel level anti cheat have never worked in proton and never will. Historically the only way to play those games is either be on windows or use a vfio virtual machine (which also probably won't work even with tons of vm hiding techniques depends on just how sensitive the ac is) . Wine/proton simply can't translate the upper parts of the nt kernel needed for it.
If the anti cheat is working in wine/proton it's not kernel level