3 ms·

A reimplementation of NetBSD using a MicroKernel (2015)

64 points by AdmiralAsshat 9y ago

carussell 9y ago
(2015) Previously: https://news.ycombinator.com/item?id=9735358 https://news.ycombinator.com/item?id=9735358 And see [1], my note about this talk being more or less the same as the 2010 Minix talk at FOSDEM[2]. If you saw that one, or any other talk by Tanenbaum of that period, I'll spoil it for you: there's nothing substantially different in this talk. 1. https://news.ycombinator.com/item?id=9736214 https://news.ycombinator.com/item?id=9736214 2. https://www.youtube.com/watch?v=bx3KuE7UjGA https://www.youtube.com/watch?v=bx3KuE7UjGA
- AdmiralAsshat 9y ago
  Apologies, I checked the date and the comments before submitting and they looked recent. I wasn't aware that the article itself was a repost. Clearly I should've checked the date on the embedded Youtube video!
  - smcl 9y ago
    I hadn't seen it before, and HN wouldn't have let you submit it if the link had been posted too much recently. So thanks for sharing! :-D
bch 9y ago
Now we just need Minix to start approaching the hardware portability that NetBSD has fought so hard for.
- snvzz 9y ago
  Those cases are well covered by netbsd, which runs well even on some Amiga 1200 with an 68030 accelerator board I've got. I've tested recent Minix3 release candidates. Software compatibility is starting to get there, as unlike previous versions, a good portion of pkgsrc builds and runs, including X11 software. However, hardware support is not very good even on x86. It boots fine, X11 can be used well with VESA drivers and most popular NICs do work, but USB is missing, which means keyboard/mouse needs to be attached to PS2 ports. As I understand it, there's USB support, but only for the beaglebone board. I know there's an effort to make it work on x86 too, but it won't be in this release. The main hardship Minix3 is currently experience is that they've been releasing RCs for 3.4 since early last year, and the actual 3.4 is still not out. A lot of work has been done which won't be on 3.4, and the delays on this release are effectively delaying everything else. They need to focus their efforts on closing the blockers, release, and move to some more frequent release schedule, such as strictly time-based; if a feature isn't ready, then it gets delayed to the next release, but releases do need to keep happening. There's a Minixcon soon; I'm hoping it triggers some progress to happen. The project is this close to the important landmark which is being somewhat usable as a workstation. After that, both press and development would pick up considerably.
  - bch 9y ago
    Well, according to http://www.minix3.org http://www.minix3.org, MINIXCon2017 is cancelled due to shortage of papers, so there won't be progress-triggering happening there... I hope you're right (and think that you are) that getting itself sort of "self hosting" as a development/workstation will increase the acceleration of its progress. edit: s/minux/minix/
    - snvzz 9y ago
      Bummer. That's extremely sad. They really do need to get that release out, and move on.
butterisgood 9y ago
Linux can be upgraded without a full reboot using the now-somewhat-ancient two-kernel-monte, or kexec? Plan 9 had been able to load new kernels and such without rebooting the machine fully as well. I love when people make crazy claims like "NO ONE ELSE CAN DO THIS... I MEAN WE NEVER ACTUALLY VERIFIED THIS OR LOOKED BUT WE'RE PRETTY SURE NO ONE ELSE CAN DO THIS!" Should go into marketing where that crap is allowed.
- butterisgood 9y ago
  Or at least explain how it's different from the existing things that work pretty well.
- john75 9y ago
  i have to admin that im not totally familiar with the internal works of kexec, but at least from my knowledge a full service restart is required[0]. The Minix team claims the upgrade happens "without running processes even noticing", that sounds better than restarting all the services. [0] https://ma.ttias.be/kernel-patching-kexec-updating-centos-7-kernel-without-full-reboot/ https://ma.ttias.be/kernel-patching-kexec-updating-centos-7-...
- snvzz 9y ago
  >Linux can be upgraded without a full reboot using the now-somewhat-ancient two-kernel-monte, or kexec? Sure, and some time is saved, as the firmware/bootloader doesn't need to run. But otherwise, like with a reboot, everything is shut down and restarted. Minix3 can update system servers and drivers in place, without rebooting and without user processes even feeling a hiccup. It can even go back to the older, working version if there's any issue with the new one. >I love when people make crazy claims like "NO ONE ELSE CAN DO THIS... I MEAN WE NEVER ACTUALLY VERIFIED THIS OR LOOKED BUT WE'RE PRETTY SURE NO ONE ELSE CAN DO THIS!" And I'm still waiting to see such an example. Linux's kexec is not.
- zlynx 9y ago
  Linux can use kpatch to upgrade the kernel while it is running. The difficult part is writing and validating the actual patch. I recall that the only people writing kpatches charge a lot of money for them and only provide them to their paying support customers. Then all of the free software people complain loudly. Yet, you don't see THEM writing (and TESTING) these functions to update in-kernel data structures in place.

64 points by AdmiralAsshat 9y ago

(2015) Previously: https://news.ycombinator.com/item?id=9735358 https://news.ycombinator.com/item?id=9735358 And see [1], my note about this talk being more or less the same as the 2010 Minix talk at FOSDEM[2]. If you saw that one, or any other talk by Tanenbaum of that period, I'll spoil it for you: there's nothing substantially different in this talk. 1. https://news.ycombinator.com/item?id=9736214 https://news.ycombinator.com/item?id=9736214 2. https://www.youtube.com/watch?v=bx3KuE7UjGA https://www.youtube.com/watch?v=bx3KuE7UjGA

AdmiralAsshat 9y ago
Apologies, I checked the date and the comments before submitting and they looked recent. I wasn't aware that the article itself was a repost. Clearly I should've checked the date on the embedded Youtube video!
- smcl 9y ago
  I hadn't seen it before, and HN wouldn't have let you submit it if the link had been posted too much recently. So thanks for sharing! :-D

bch 9y ago

Now we just need Minix to start approaching the hardware portability that NetBSD has fought so hard for.

snvzz 9y ago
Those cases are well covered by netbsd, which runs well even on some Amiga 1200 with an 68030 accelerator board I've got. I've tested recent Minix3 release candidates. Software compatibility is starting to get there, as unlike previous versions, a good portion of pkgsrc builds and runs, including X11 software. However, hardware support is not very good even on x86. It boots fine, X11 can be used well with VESA drivers and most popular NICs do work, but USB is missing, which means keyboard/mouse needs to be attached to PS2 ports. As I understand it, there's USB support, but only for the beaglebone board. I know there's an effort to make it work on x86 too, but it won't be in this release. The main hardship Minix3 is currently experience is that they've been releasing RCs for 3.4 since early last year, and the actual 3.4 is still not out. A lot of work has been done which won't be on 3.4, and the delays on this release are effectively delaying everything else. They need to focus their efforts on closing the blockers, release, and move to some more frequent release schedule, such as strictly time-based; if a feature isn't ready, then it gets delayed to the next release, but releases do need to keep happening. There's a Minixcon soon; I'm hoping it triggers some progress to happen. The project is this close to the important landmark which is being somewhat usable as a workstation. After that, both press and development would pick up considerably.
- bch 9y ago
  Well, according to http://www.minix3.org http://www.minix3.org, MINIXCon2017 is cancelled due to shortage of papers, so there won't be progress-triggering happening there... I hope you're right (and think that you are) that getting itself sort of "self hosting" as a development/workstation will increase the acceleration of its progress. edit: s/minux/minix/
  - snvzz 9y ago
    Bummer. That's extremely sad. They really do need to get that release out, and move on.

butterisgood 9y ago

Linux can be upgraded without a full reboot using the now-somewhat-ancient two-kernel-monte, or kexec? Plan 9 had been able to load new kernels and such without rebooting the machine fully as well. I love when people make crazy claims like "NO ONE ELSE CAN DO THIS... I MEAN WE NEVER ACTUALLY VERIFIED THIS OR LOOKED BUT WE'RE PRETTY SURE NO ONE ELSE CAN DO THIS!" Should go into marketing where that crap is allowed.

butterisgood 9y ago
Or at least explain how it's different from the existing things that work pretty well.
john75 9y ago
i have to admin that im not totally familiar with the internal works of kexec, but at least from my knowledge a full service restart is required[0]. The Minix team claims the upgrade happens "without running processes even noticing", that sounds better than restarting all the services. [0] https://ma.ttias.be/kernel-patching-kexec-updating-centos-7-kernel-without-full-reboot/ https://ma.ttias.be/kernel-patching-kexec-updating-centos-7-...
snvzz 9y ago
>Linux can be upgraded without a full reboot using the now-somewhat-ancient two-kernel-monte, or kexec? Sure, and some time is saved, as the firmware/bootloader doesn't need to run. But otherwise, like with a reboot, everything is shut down and restarted. Minix3 can update system servers and drivers in place, without rebooting and without user processes even feeling a hiccup. It can even go back to the older, working version if there's any issue with the new one. >I love when people make crazy claims like "NO ONE ELSE CAN DO THIS... I MEAN WE NEVER ACTUALLY VERIFIED THIS OR LOOKED BUT WE'RE PRETTY SURE NO ONE ELSE CAN DO THIS!" And I'm still waiting to see such an example. Linux's kexec is not.
zlynx 9y ago
Linux can use kpatch to upgrade the kernel while it is running. The difficult part is writing and validating the actual patch. I recall that the only people writing kpatches charge a lot of money for them and only provide them to their paying support customers. Then all of the free software people complain loudly. Yet, you don't see THEM writing (and TESTING) these functions to update in-kernel data structures in place.