IRC channel logs

2022-01-16.log

back to list of logs

<Gooberpatrol66>could netbsd's compat_linux be used in conjunction with rump to bring linux binary compatibility to hurd?
<Curiosa>Gooberpatrol66: it would be more compatible and probably feasible to port the Linux Kernel to the Hurd as a single server running as a user process
<Curiosa>Not very different from supporting it under qemu
<Curiosa>But more similar perhaps to what windows does? Dunno the details of that
<Curiosa>Well, better then qemu since Linux processes could map to Hurd processes and use that server for syscalls
<yjftsjthsd3>User Mode Linux is a thing that already exists; AFAIK it only runs on Linux right now, but if you wanted to port Linux to HURD as a user process that's probably the easiest starting point
<yjftsjthsd3>Alternatively, start from NetBSD's Linux compat and implement just enough to run UML for full compatibility:) (Or maybe that's harder; I'm not a kernel dev:D)
<luckyluke>lkl (lkl.github.io) would be a better approach if it would progress, it could be used to provide directly the syscalls, e.g. to use the network stack, or write drivers
<luckyluke>unfortunately it seems to have stalled
<youpi>I'm not really surprised. There has been various projects on reusing the Linux kernel source, but since its internal API is very moving, it's hard to follow
<youpi>targetting netbsd is much more sustainable
<youpi>especially now that rump is upstreamed :)
<luckyluke>sure, the only drawback I see is that netbsd has less drivers
<luckyluke>I had LKL mostly running on GNU/Hurd a few years ago, but I remember I had issues with irqs
<luckyluke>so I couldn't really reuse the driver I wanted
<luckyluke>but for filesystems and "higher level" drivers it was working
<luckyluke>I even created a .deb package at the time :)
<Curiosa>Linux has already been ported to the Mach kernel, porting it to the Hurd should be probably similarly approachable but in many way more powerful/interesting
<Curiosa> https://www.tech-insider.org/linux/research/acrobat/960112.pdf
<Gooberpatrol66>luckyluke: do you have a link to the code for that?
<Curiosa>Of the paper I listed Fake interrupts might be interesting for Hurd itself since last time I checked posix signal handling had little issues
<luckyluke>Gooberpatrol66: IIRC it should be here https://gitlab.com/luckyd/linux/-/tree/debian/sid
<luckyluke>build with git-buildpackage
<Gooberpatrol66>luckyluke: thx
<Curiosa>The Linux port to Mach is stuck at version 2.0, nevermind https://github.com/slp/osfmk-mklinux
<kilobug>Curiosa: it was a mix of an experiment and a way to boot Linux on some Mac hardware at a time, but yes it's not a project that has been maintained for "production use"
<Curiosa>kilobug: I see, it’s that looking at the code it seems like a lot of work, but you’re also left wondering if that was even legal from a licensing viewpoint
<Curiosa>They seem like mixing gpl2 code with source files whose license had clearly been removed along with documentation