i've watched my roommate--a person whose technical skills i respect--spend several weeks trying to set up a dual booting nixos/win10 system with uefi
i believe she's on the fifth reinstall from scratch, and having seen that breakage, i know i won't do much better, either
Conversation
Used to be in a tiny IRC channel with someone that used Xen to be able to run Windows for gaming with PCIe passthrough. 90% of their posts in that channel (not really a technical channel) was complaining about PCIe passthrough breaking and asking for help debugging it.
1
5
yeah I've tried it, hit an issue with PAT/MTRR and write combining ranges in graphical memory, tried fixing it, decided KVM it is (I might actually set up KVM with PCIe passthrough if wine doesn't work out)
1
2
I personally prefer having Windows on a totally different machine for gaming. For some reason I still have my better GPU in my workstation from back when I thought dual booting was a good idea but... that'll be fixed when I replace it with a new workstation.
2
I have a spare TB nvme waiting for me to drop linux on it in my big workstation, and I've been looking at various distros update behavior in fear. I think systemd-boot is going to be ok, but half the distros hold it wrong
1
I have quite literally been considering starting a distro just to get sane updates and reasonable systemd purity recently.
2
I've used a 2TB Samsung 960 Pro NVMe drive as the only drive in my workstation for several years with zero issues.
I just use EFISTUB and if a kernel regression breaks something, I can boot up the LTS kernel via the annoying UEFI menu.
Never had any issues with using NVMe.
1
1
What I mean is they have bad behavior if they have to share a UEFI ESP
1
1
When I used to dual boot, I had Windows on a completely separate drive with a separate ESP and I'd just tell UEFI to boot from that drive when I wanted to use Windows.
2
Yeah, that's been the recommendation I've seen given in related bug reports from most distros, and it makes me sad. There's a whole protocol for this and it's in a fairly open spec thats only a few hundred pages (most of which has nothing to do with booting)
1
1
I've successfully had a shared ESP but Windows leaves the partition in a weird state when you 'shut down' since it actually does some hybrid hibernation thing and it screws up booting Linux. Can deal with that stuff and the UTC RTC issue but... I'd rather just have it separate.
I saw that report about NTFS, I hadn't seen that being done with the ESP, but I can believe it. I'm slightly surprised windows bothers to hold it in a dirty state though, it doesn't write to it much at all.
1
I'm also skeptical that "dirty state" has any significant meaning with fat32 really, but there are always plenty of ways to fuck this up
1