9 ms·

FreeBSD 10.0 Alpha 1 now available

111 points by daw___ 13y ago

pjenvey 13y ago
What's new: https://wiki.freebsd.org/WhatsNew/FreeBSD10 https://wiki.freebsd.org/WhatsNew/FreeBSD10
lcampbell 13y ago
The What's New for FreeBSD 10 wiki page[1] linked in the reply has more details (though it doesn't necessarily indicate what will make the cut). The interesting bits to me are the merging of VPS support[2] which, last I read was a patchset against the 7 or 8 kernel. The seamless migration of virtualized envs seemed like a pipe dream at the time, I can't wait to play around with it (if it makes the cut). Not to mention virtio support and bhyve[3] (in addition to the old jail subsystem). If everything makes it in, there will be so many different options for service isolation! -- [1] https://wiki.freebsd.org/WhatsNew/FreeBSD10 https://wiki.freebsd.org/WhatsNew/FreeBSD10 [2] http://7he.at/freebsd/vps/ http://7he.at/freebsd/vps/ [3] https://wiki.freebsd.org/action/show/bhyve?action=show&redirect=BHyVe https://wiki.freebsd.org/action/show/bhyve?action=show&redir...
- gonzo 13y ago
  it's likely to make the cut, given that it's pretty damned stable. virio will be huge.
  - rwmj 13y ago
    Virtio is excellent news, and I'm glad to see they've implemented virtio-scsi which is just in every way much better than the old virtio-blk (although they've implemented that too).
- sgt 13y ago
  bhyve seems very interesting - is there any mention of IO and CPU throttling ? I want to be able to at least have rudimentary control of how many CPU cycles a VM is allowed to use, when competing with other VMs for the same CPU cycles and resources.
ubikation 13y ago
So how does bhyve compare to kvm? I've seen FreeBSD people saying it's a 'legacy free' virtualization system but I don't really know what that means or entails. Can it do live migrations and failovers like xen/kvm? What can it do that kvm/xen can't? What about vice versa?
- kev009 13y ago
  It's legacy free in the sense that it doesn't do paravirt like Xen and requires hardware assistance like KVM. Maybe also that it doesn't use QEMU which has some pros and cons: http://www.slideshare.net/syuu1228/implements-bios-emulation-support-for-b-hy-ve-a-bsd-hypervisor http://www.slideshare.net/syuu1228/implements-bios-emulation.... Ultimately, it will let people like NetApp use it on their filers without any GPL complications though. In all other respects, it's still in its infancy and I wouldn't expect KVM parity for common use cases until the next release. I'm more excited about the VIMAGE stuff in 10.
synchronise 13y ago
Personally I can't wait to use 10 when it gets released. But it's a pity that it won't see Gnome 3.10 due to its dependence on logind, and then logind's new dependency on systemd as of 205.
a--b 13y ago
I've had a long standing question about FreeBSD and ports/portsnap: does/how does FreeBSD prevent against bad mirrors or MITM attacks?
- guerrilla 13y ago
  http://forums.freebsd.org/archive/index.php/t-27111.html http://forums.freebsd.org/archive/index.php/t-27111.html
- gaadd33 13y ago
  AFAIK ports checks the checksum of the file when it downloads it against the checksums the ports maintainer submitted. For the ISO mirrors, I don't think there's anything automatic, you have to manually run sha256 on the file and compare checksums. Do Linux distributions have another way?
  - adestefan 13y ago
    Most Linux package managers use GPG signed packages.
    - gaadd33 13y ago
      Right but if you are downloading an ISO you still have to manually verify it correct?
      - lifeisstillgood 13y ago
        yes but a one time md5 of an iso is far less strenuous than manually checking the provenance of 200 tarballs to build say X. there is a new package management system similar to apt-get but it's far rarer to see it used than ports I really like ports - as much for the knowledge that I have all the right header files.
        gaadd33 13y ago
        When do you have to check the provenance of 200 tarballs to install X? If you install from an ISO, you can install the ports from there too and you have checksums which are automatically checked, right? So if you do cd /usr/ports/x11-wm/xfce4; make install it will automatically check all of the checksums right?
erichurkman 13y ago
It sounds like some amazing performance boosts around networking. I wonder how much of that will carry over to all of the network-type appliances built on FreeBSD? I know Juno's OS runs a pretty ancient version of FreeBSD (6 or 7 IIRC), but am not sure about any other commercial uses.
- profquail 13y ago
  The `netmap` stuff (high-performance networking stack) is really neat. The author's site has a lot more information about it if you're interested: http://info.iet.unipi.it/~luigi/netmap/ http://info.iet.unipi.it/~luigi/netmap/ I'm curious to see what kind of performance is possible with FreeBSD 10 (using `netmap`) and the new I/O manager which will be included in the upcoming GHC 7.8 release. The combination seems like it would be epic in terms of web server performance, and it might be a nice choice for web hosting companies / CDNs who have a vested interest in reducing latency as much as possible.
  - wmf 13y ago
    First somebody needs to create a TCP stack for Netmap; right now it only deals in Ethernet packets. Netmap has the potential for very high performance; we're capturing 50 Gbps and hoping to hit 140 Gbps per server soon.
- pktgen 13y ago
  Hmm, it might make some difference on some of the Juniper lines (is the SRX line software-based?) but any of the high-end Juniper products are only using JunOS for the control plane. Only control traffic and certain exception cases (e.g. ICMP) ever hit the BSD/JunOS network stack; the rest is all handled in hardware.
lifeisstillgood 13y ago
quick summary for the impatient: - drop GCC for CLANG. - introduce VPS (more virtualised approach to jails) - introduce bhyve (native hypervisor on common standards) - replace BIND - Can be run on raspberry pi - major speed ups in SMP, and networking FreeBSD Is taking a fairly big leap forward in its sweet spot of the workhorse of the data farm. BSD has always had great jails virtualisation support (it's what LXC or docker is following) but now that's being expanded and complemented with the run-another-OS-on-my-host-OS of "virtio" style VMs Raspberry pi of course is what the world has been waiting for - never mind all the rest :-) edit: tidy ups
- profquail 13y ago
  FreeBSD 10 is also going to have much-improved support for Xen [1][2] and newly-added support for Hyper-V [3]. I'd love to see AWS and Azure support first-class FreeBSD instances. (You can currently run FreeBSD instances using cperciva's AMIs [4], but that's not the same as Amazon supporting FreeBSD in the same way they support Linux). [1]: http://lists.freebsd.org/pipermail/freebsd-xen/2013-June/001618.html http://lists.freebsd.org/pipermail/freebsd-xen/2013-June/001... [2]: http://wiki.xen.org/wiki/Testing_FreeBSD_PVHVM http://wiki.xen.org/wiki/Testing_FreeBSD_PVHVM [3]: http://forums.freebsd.org/showthread.php?t=38019 http://forums.freebsd.org/showthread.php?t=38019 [4]: http://www.daemonology.net/freebsd-on-ec2/ http://www.daemonology.net/freebsd-on-ec2/
  - lifeisstillgood 13y ago
    agreed with AWS and freebsd. although there is hardly the demand - each year I have a debate whether to forego BSD entirely (especially on the desktop) but its just that too good to lose. it looks like I'm going to get saltstack working with VpS and keep with Rackspace :-)
    - profquail 13y ago
      > although there is hardly the demand It's a chicken/egg problem -- Amazon won't be motivated to implement additional support for FreeBSD without more people asking for it, but the additional visibility FreeBSD would gain from better support on AWS (and/or Azure) would definitely increase the number of people using it. It seems like FreeBSD has seen an upswing in interest over the past couple of years, and if interest keeps growing I think it'd eventually hit some threshold where Amazon is willing to implement better support for FreeBSD and then put some marketing behind it. On FreeBSD + saltstack: I saw an interesting article a few weeks back (on the saltstack website, no less) about them implementing better support for FreeBSD, specifically w.r.t. jails: http://intothesaltmine.org/blog/html/2013/08/30/bootstrapping_freebsd_jails_with_salt_bootstrap.html http://intothesaltmine.org/blog/html/2013/08/30/bootstrappin...
kunai 13y ago
I run BSD on all of my desktop boxes, but I have not found any way to run it on a laptop. It just doesn't seem feasible; there are just so many mobile drivers that are missing. Has anyone else had any luck with getting FreeBSD to work on a laptop, or is Linux still the de-facto?
- eulerphi 13y ago
  I had FreeBSD running on my old Dell Studio 1737. It took some effort but most of the linux drivers will work with BSD if you edit the source and recompile.
  - kunai 13y ago
    Any links? If not, a short anecdote? How is power management and suspend?
  - 4ad 13y ago
    You most definitely can't recompile Linux drivers for FreeBSD. The driver model is very kernel specific, porting drivers is more close to rewriting drivers. Plus there's the GPL thing. Linux is GPL, that's why you'll never see Linux drivers ported to BSDs. Sys V drivers conforming to the STREAMS model were supposed to be portable between Unix variants, but STREAMS never catched on, neither Linux nor BSDs use STREAMS, and BSDs only have minimal STREAMS support for the SVR4 compatibility layer. Solaris has STREAMS, but has rewritten many drivers not to use the, e.g. the networking stack has been rewritten from being STREAM-based (now it's called Fire Engine). It's generally easier to port BSD drivers to Solaris, then it is to port BSD drivers to Linux.
    - belorn 13y ago
      I do find it odd that when BSD people talk about the BSD license, they describe it as pragmatism winning over idealism. However, when they talk about the BSD OS's, suddenly that pragmatism disappear and everything need to be pure permissive licensed. Is both "camps" just similar fanatic regarding licensing, similar pragmatic, or is it a natural distinct difference when talking about licenses, and when picking software licensed under one of the two types?
rwmj 13y ago
Can anyone tell me more about the hypervisor (http://bhyve.org/ http://bhyve.org/)? The design looks extremely similar to KVM, so I don't understand why they didn't just implement the KVM ioctls, on the basis that they would be able to run all the existing userspace (ie. qemu). The problem with implementing a similar-but-different kernel API is that you've either got to persuade qemu to implement your API as well as KVM, or you have to rewrite the whole of qemu (a massive job).
- 4ad 13y ago
  They don't have to convince anyone, nor rewrite anything. It's a different architecture. Note that not even KVM doesn't technically need Qemu in particular. Qemu is overkill for most usage of KVM, but they used it because it was available. Some people wrote something much smaller that can be used with KVM: https://lkml.org/lkml/2011/3/31/406 https://lkml.org/lkml/2011/3/31/406
  - rwmj 13y ago
    Sure, except kvmtool was never accepted upstream and has since been abandoned. To make kvmtool useful would have involved tediously reimplementing large parts of qemu/seabios (qcow2 support, hundreds of emulated devices, firmware). Which is the same thing that FreeBSD will have to do, if they want to run a mix of real world guests any time soon. Pure programming is great and I've no problem if FreeBSD wants to take years and implement something fantastic.
    - 4ad 13y ago
      Emulating real hardware is pointless if your use case is running Linux/FreeBSD in production[1]. It's only important if you want to run historic operating systems or Windows[2]. Effort is more usefully directed into making the hypervisor better rather than implementing legacy cruft. Why should I emulate An A20 gate if my current Linux kernels do just fine without it? [1] Other people seem to agree with this as they were willing to use Xen before the advent of CPUs which offered hardware assisted virtualization. Xen required a special paravirtualized Linux kernel and people had no problem with this. [2] In my experience people who want to run Windows use Hyper-V or VMware, not Xen/KVM.
telephonetemp 13y ago
Is there information as to whether pkgng will be available in time for the 10.0 release?
- profquail 13y ago
  AFAIK, pkgng is the default package manager in both FreeBSD 9.2 (due any day now) and 10.0. The table at the bottom of this page says so anyway, and it was updated just a few months back: https://wiki.freebsd.org/pkgng/CharterAndRoadMap https://wiki.freebsd.org/pkgng/CharterAndRoadMap
freeman478 13y ago
Dffcf-123. ,l,km

111 points by daw___ 13y ago

What's new: https://wiki.freebsd.org/WhatsNew/FreeBSD10 https://wiki.freebsd.org/WhatsNew/FreeBSD10

The What's New for FreeBSD 10 wiki page[1] linked in the reply has more details (though it doesn't necessarily indicate what will make the cut). The interesting bits to me are the merging of VPS support[2] which, last I read was a patchset against the 7 or 8 kernel. The seamless migration of virtualized envs seemed like a pipe dream at the time, I can't wait to play around with it (if it makes the cut). Not to mention virtio support and bhyve[3] (in addition to the old jail subsystem). If everything makes it in, there will be so many different options for service isolation! -- [1] https://wiki.freebsd.org/WhatsNew/FreeBSD10 https://wiki.freebsd.org/WhatsNew/FreeBSD10 [2] http://7he.at/freebsd/vps/ http://7he.at/freebsd/vps/ [3] https://wiki.freebsd.org/action/show/bhyve?action=show&redirect=BHyVe https://wiki.freebsd.org/action/show/bhyve?action=show&redir...

gonzo 13y ago
it's likely to make the cut, given that it's pretty damned stable. virio will be huge.
- rwmj 13y ago
  Virtio is excellent news, and I'm glad to see they've implemented virtio-scsi which is just in every way much better than the old virtio-blk (although they've implemented that too).
sgt 13y ago
bhyve seems very interesting - is there any mention of IO and CPU throttling ? I want to be able to at least have rudimentary control of how many CPU cycles a VM is allowed to use, when competing with other VMs for the same CPU cycles and resources.

ubikation 13y ago

So how does bhyve compare to kvm? I've seen FreeBSD people saying it's a 'legacy free' virtualization system but I don't really know what that means or entails. Can it do live migrations and failovers like xen/kvm? What can it do that kvm/xen can't? What about vice versa?

kev009 13y ago
It's legacy free in the sense that it doesn't do paravirt like Xen and requires hardware assistance like KVM. Maybe also that it doesn't use QEMU which has some pros and cons: http://www.slideshare.net/syuu1228/implements-bios-emulation-support-for-b-hy-ve-a-bsd-hypervisor http://www.slideshare.net/syuu1228/implements-bios-emulation.... Ultimately, it will let people like NetApp use it on their filers without any GPL complications though. In all other respects, it's still in its infancy and I wouldn't expect KVM parity for common use cases until the next release. I'm more excited about the VIMAGE stuff in 10.

synchronise 13y ago

Personally I can't wait to use 10 when it gets released. But it's a pity that it won't see Gnome 3.10 due to its dependence on logind, and then logind's new dependency on systemd as of 205.

a--b 13y ago

I've had a long standing question about FreeBSD and ports/portsnap: does/how does FreeBSD prevent against bad mirrors or MITM attacks?

guerrilla 13y ago
http://forums.freebsd.org/archive/index.php/t-27111.html http://forums.freebsd.org/archive/index.php/t-27111.html
gaadd33 13y ago
AFAIK ports checks the checksum of the file when it downloads it against the checksums the ports maintainer submitted. For the ISO mirrors, I don't think there's anything automatic, you have to manually run sha256 on the file and compare checksums. Do Linux distributions have another way?
- adestefan 13y ago
  Most Linux package managers use GPG signed packages.
  - gaadd33 13y ago
    Right but if you are downloading an ISO you still have to manually verify it correct?
    - lifeisstillgood 13y ago
      yes but a one time md5 of an iso is far less strenuous than manually checking the provenance of 200 tarballs to build say X. there is a new package management system similar to apt-get but it's far rarer to see it used than ports I really like ports - as much for the knowledge that I have all the right header files.
      - gaadd33 13y ago
        When do you have to check the provenance of 200 tarballs to install X? If you install from an ISO, you can install the ports from there too and you have checksums which are automatically checked, right? So if you do cd /usr/ports/x11-wm/xfce4; make install it will automatically check all of the checksums right?

erichurkman 13y ago

It sounds like some amazing performance boosts around networking. I wonder how much of that will carry over to all of the network-type appliances built on FreeBSD? I know Juno's OS runs a pretty ancient version of FreeBSD (6 or 7 IIRC), but am not sure about any other commercial uses.

profquail 13y ago
The `netmap` stuff (high-performance networking stack) is really neat. The author's site has a lot more information about it if you're interested: http://info.iet.unipi.it/~luigi/netmap/ http://info.iet.unipi.it/~luigi/netmap/ I'm curious to see what kind of performance is possible with FreeBSD 10 (using `netmap`) and the new I/O manager which will be included in the upcoming GHC 7.8 release. The combination seems like it would be epic in terms of web server performance, and it might be a nice choice for web hosting companies / CDNs who have a vested interest in reducing latency as much as possible.
- wmf 13y ago
  First somebody needs to create a TCP stack for Netmap; right now it only deals in Ethernet packets. Netmap has the potential for very high performance; we're capturing 50 Gbps and hoping to hit 140 Gbps per server soon.
pktgen 13y ago
Hmm, it might make some difference on some of the Juniper lines (is the SRX line software-based?) but any of the high-end Juniper products are only using JunOS for the control plane. Only control traffic and certain exception cases (e.g. ICMP) ever hit the BSD/JunOS network stack; the rest is all handled in hardware.

lifeisstillgood 13y ago

quick summary for the impatient: - drop GCC for CLANG. - introduce VPS (more virtualised approach to jails) - introduce bhyve (native hypervisor on common standards) - replace BIND - Can be run on raspberry pi - major speed ups in SMP, and networking FreeBSD Is taking a fairly big leap forward in its sweet spot of the workhorse of the data farm. BSD has always had great jails virtualisation support (it's what LXC or docker is following) but now that's being expanded and complemented with the run-another-OS-on-my-host-OS of "virtio" style VMs Raspberry pi of course is what the world has been waiting for - never mind all the rest :-) edit: tidy ups

profquail 13y ago
FreeBSD 10 is also going to have much-improved support for Xen [1][2] and newly-added support for Hyper-V [3]. I'd love to see AWS and Azure support first-class FreeBSD instances. (You can currently run FreeBSD instances using cperciva's AMIs [4], but that's not the same as Amazon supporting FreeBSD in the same way they support Linux). [1]: http://lists.freebsd.org/pipermail/freebsd-xen/2013-June/001618.html http://lists.freebsd.org/pipermail/freebsd-xen/2013-June/001... [2]: http://wiki.xen.org/wiki/Testing_FreeBSD_PVHVM http://wiki.xen.org/wiki/Testing_FreeBSD_PVHVM [3]: http://forums.freebsd.org/showthread.php?t=38019 http://forums.freebsd.org/showthread.php?t=38019 [4]: http://www.daemonology.net/freebsd-on-ec2/ http://www.daemonology.net/freebsd-on-ec2/
- lifeisstillgood 13y ago
  agreed with AWS and freebsd. although there is hardly the demand - each year I have a debate whether to forego BSD entirely (especially on the desktop) but its just that too good to lose. it looks like I'm going to get saltstack working with VpS and keep with Rackspace :-)
  - profquail 13y ago
    > although there is hardly the demand It's a chicken/egg problem -- Amazon won't be motivated to implement additional support for FreeBSD without more people asking for it, but the additional visibility FreeBSD would gain from better support on AWS (and/or Azure) would definitely increase the number of people using it. It seems like FreeBSD has seen an upswing in interest over the past couple of years, and if interest keeps growing I think it'd eventually hit some threshold where Amazon is willing to implement better support for FreeBSD and then put some marketing behind it. On FreeBSD + saltstack: I saw an interesting article a few weeks back (on the saltstack website, no less) about them implementing better support for FreeBSD, specifically w.r.t. jails: http://intothesaltmine.org/blog/html/2013/08/30/bootstrapping_freebsd_jails_with_salt_bootstrap.html http://intothesaltmine.org/blog/html/2013/08/30/bootstrappin...

kunai 13y ago

I run BSD on all of my desktop boxes, but I have not found any way to run it on a laptop. It just doesn't seem feasible; there are just so many mobile drivers that are missing. Has anyone else had any luck with getting FreeBSD to work on a laptop, or is Linux still the de-facto?

eulerphi 13y ago
I had FreeBSD running on my old Dell Studio 1737. It took some effort but most of the linux drivers will work with BSD if you edit the source and recompile.
- kunai 13y ago
  Any links? If not, a short anecdote? How is power management and suspend?
- 4ad 13y ago
  You most definitely can't recompile Linux drivers for FreeBSD. The driver model is very kernel specific, porting drivers is more close to rewriting drivers. Plus there's the GPL thing. Linux is GPL, that's why you'll never see Linux drivers ported to BSDs. Sys V drivers conforming to the STREAMS model were supposed to be portable between Unix variants, but STREAMS never catched on, neither Linux nor BSDs use STREAMS, and BSDs only have minimal STREAMS support for the SVR4 compatibility layer. Solaris has STREAMS, but has rewritten many drivers not to use the, e.g. the networking stack has been rewritten from being STREAM-based (now it's called Fire Engine). It's generally easier to port BSD drivers to Solaris, then it is to port BSD drivers to Linux.
  - belorn 13y ago
    I do find it odd that when BSD people talk about the BSD license, they describe it as pragmatism winning over idealism. However, when they talk about the BSD OS's, suddenly that pragmatism disappear and everything need to be pure permissive licensed. Is both "camps" just similar fanatic regarding licensing, similar pragmatic, or is it a natural distinct difference when talking about licenses, and when picking software licensed under one of the two types?

rwmj 13y ago

Can anyone tell me more about the hypervisor (http://bhyve.org/ http://bhyve.org/)? The design looks extremely similar to KVM, so I don't understand why they didn't just implement the KVM ioctls, on the basis that they would be able to run all the existing userspace (ie. qemu). The problem with implementing a similar-but-different kernel API is that you've either got to persuade qemu to implement your API as well as KVM, or you have to rewrite the whole of qemu (a massive job).

4ad 13y ago
They don't have to convince anyone, nor rewrite anything. It's a different architecture. Note that not even KVM doesn't technically need Qemu in particular. Qemu is overkill for most usage of KVM, but they used it because it was available. Some people wrote something much smaller that can be used with KVM: https://lkml.org/lkml/2011/3/31/406 https://lkml.org/lkml/2011/3/31/406
- rwmj 13y ago
  Sure, except kvmtool was never accepted upstream and has since been abandoned. To make kvmtool useful would have involved tediously reimplementing large parts of qemu/seabios (qcow2 support, hundreds of emulated devices, firmware). Which is the same thing that FreeBSD will have to do, if they want to run a mix of real world guests any time soon. Pure programming is great and I've no problem if FreeBSD wants to take years and implement something fantastic.
  - 4ad 13y ago
    Emulating real hardware is pointless if your use case is running Linux/FreeBSD in production[1]. It's only important if you want to run historic operating systems or Windows[2]. Effort is more usefully directed into making the hypervisor better rather than implementing legacy cruft. Why should I emulate An A20 gate if my current Linux kernels do just fine without it? [1] Other people seem to agree with this as they were willing to use Xen before the advent of CPUs which offered hardware assisted virtualization. Xen required a special paravirtualized Linux kernel and people had no problem with this. [2] In my experience people who want to run Windows use Hyper-V or VMware, not Xen/KVM.

telephonetemp 13y ago

Is there information as to whether pkgng will be available in time for the 10.0 release?

profquail 13y ago
AFAIK, pkgng is the default package manager in both FreeBSD 9.2 (due any day now) and 10.0. The table at the bottom of this page says so anyway, and it was updated just a few months back: https://wiki.freebsd.org/pkgng/CharterAndRoadMap https://wiki.freebsd.org/pkgng/CharterAndRoadMap

freeman478 13y ago

Dffcf-123. ,l,km