3 ms·
I agree with the skepticism here regarding the ease of getting Windows software working on Linux. The Wine and CrossOver folks are doing the Lord's work but eve
by jchw 8y ago
I agree with the skepticism here regarding the ease of getting Windows software working on Linux. The Wine and CrossOver folks are doing the Lord's work but even they have their limits.
Linux software tends to not rely on much. Linux makes it really hard to rely on much. The internals change often and the kernel is not nice to people who try to rely on implementation details, of what you can from usermode. You can hardly rely on libc when you are on Linux, and many try not to, to be more portable.
Windows software on the other hand is wild. Just look at the myriad of techniques used by anti-debugging and anti-reverse-engineering tools. A Linux binary wouldn't dream of reimplementing the runtime linker itself, but that's exactly what many packers do on Windows, to obscure the import address table and make patching/debugging harder. Did you know you can write into another processes address space with WriteProcessMemory? Why do we even have that lever!?
That's only considering usermode. But apps are just as eager to rely on kernel mode implementation details too, in the past it was even common to patch the SSDT to modify syscall behavior. Anti-cheat in video games can still do evil things even on Windows 10; nProtect GameGuard's kernel module seems to hide its usermode processes somehow. I'm pretty sure Linux kernel modules can't easily do that.
I'd love better Windows compatibility on Linux. Heck, I'm excited by what Valve is doing with Proton too. But in the end, I think much of the Windows software library is just too deeply ingrained in the Windows legacy.
- pritambaral 8y ago> I'm pretty sure Linux kernel modules can't easily do that. They can, and rootkits in the past (unsurprisingly) have. But that's just technical ability; users willing to run things like that is a whole different matter, especially when we can just namespace and/or sandbox apps.
- jchw 8y agoI'm not so surprised that rootkits can exist on Linux, but I was under the impression the Linux kernel made it much harder to break out of the framework provided and patch over whatever you wanted. Sure, it's all Ring 0, but there's kASLR and you have limited visibility, maybe not even a map of symbols. Is it still possible to do this kind of attack today?
- pritambaral 8y agoEverything running in kernel-space is fully trusted (talking about code compiled and linked against the kernel, not JITted code running in a VM). Symbol map may or may not be built and/or exported during build (both defaulting to true on Debian/Ubuntu setups).
- CoolGuySteve 8y agoWine is a lot different from virtualization. It’s likely if Microsoft did release a Windows compatible Linux (or some micro kernel thing) they would virtualize the SxS assemblies and kernel for each application to help with security and compatibility. They gave a demo called MinWin years ago that more or less did the same thing.
- jchw 8y agoThat would be a very neat approach. I had not heard of MinWin until today and now I'm pretty curious exactly what it entails and what the implications of it are.
- CoolGuySteve 8y agoIt’s actually difficult to piece together what MinWin was exactly. Some sources say it used virtualization, other source say it was an extremely limited NT kernel. I suspect it was a little of both.