Xen is a hypervisor (or virtual machine monitor) for x86 hardware (i686-class or higher), which supports running multiple guest operating systems on a single physical machine. Xen is a Type 1 or bare-metal hypervisor; one uses the Xen kernel to control the CPU, memory and console, a dom0 operating system which mediates access to other hardware (e.g., disks, network, USB), and one or more domU operating systems which operate in an unprivileged virtualized environment. IO requests from the domU systems are forwarded by the hypervisor (Xen) to the dom0 to be fulfilled.
Xen supports two styles of guests. The original is Para-Virtualized (PV) which means that the guest OS does not attempt to access hardware directly, but instead makes hypercalls to the hypervisor. This is analogous to a user-space program making system calls. (The dom0 operating system uses PV calls for some functions, such as updating memory mapping page tables, but has direct hardware access for disk and network.) PV guests must be specifically coded for Xen.
The more recent style is HVM, which means that the guest does not have code for Xen and need not be aware that it is running under Xen. Attempts to access hardware registers are trapped and emulated. This style is less efficient but can run unmodified guests.
Generally any machine that runs NetBSD/amd64 will work with Xen and PV guests. In theory i386 computers (without x86_64/amd64 support) can be used for Xen <= 4.2, but we have no recent reports of this working (this is a hint). For HVM guests, hardware support is needed, but it is common on recent machines. For Intel CPUs, one needs the VT-x extension, shown in "cpuctl identify 0" as VMX. For AMD CPUs, one needs the AMD-V extensions, shown in "cpuctl identify 0" as SVM. There are further features for IOMMU virtualization, Intel's VT-d and AMD's AMD-Vi. TODO: Explain whether Xen on NetBSD makes use of these features. TODO: Review by someone who really understands this.
Note that a FreeBSD dom0 requires VT-x and VT-d (or equivalent); this is because the FreeBSD dom0 does not run in PV mode.
At boot, the dom0 kernel is loaded as a module with Xen as the kernel. The dom0 can start one or more domUs. (Booting is explained in detail in the dom0 section.)
NetBSD supports Xen in that it can serve as dom0, be used as a domU, and that Xen kernels and tools are available in pkgsrc. This HOWTO attempts to address both the case of running a NetBSD dom0 on hardware and running domUs under it (NetBSD and other), and also running NetBSD as a domU in a VPS.
Xen 3.1 in pkgsrc supports "PCI passthrough", which means that specific PCI devices can be made available to a specific domU instead of the dom0. This can be useful to let a domU run X11, or access some network interface or other peripheral.
NetBSD 6 and earlier supported Xen 2; support was removed from NetBSD 7. Xen 2 has been removed from pkgsrc.
Installing NetBSD/Xen is not extremely difficult, but it is more complex than a normal installation of NetBSD. In general, this HOWTO is occasionally overly restrictive about how things must be done, guiding the reader to stay on the established path when there are no known good reasons to stray.
This HOWTO presumes a basic familiarity with the Xen system architecture, with installing NetBSD on i386/amd64 hardware, and with installing software from pkgsrc. See also the Xen website.
Versions of Xen and NetBSD
Most of the installation concepts and instructions are independent of Xen version and NetBSD version. This section gives advice on which version to choose. Versions not in pkgsrc and older unsupported versions of NetBSD are intentionally ignored.
The term "amd64" is used to refer to both the NetBSD port and to the hardware architecture on which it runs. (Such hardware is made by both Intel and AMD, and in 2016 a normal PC has this CPU architecture.)
In NetBSD, Xen is provided in pkgsrc, via matching pairs of packages xenkernel and xentools. We will refer only to the kernel versions, but note that both packages must be installed together and must have matching versions.
xenkernel3 provides Xen 3.1. It is no longer maintained by Xen, and the last applied security patch was in 2011. Thus, it should not be used. It supports PCI passthrough, which is why people use it anyway. Xen 3.1 runs on i386 (both non-PAE and PAE) and amd64 hardware.
xenkernel33 provides Xen 3.3. It is no longer maintained by Xen, and the last applied security patch was in 2012. Thus, it should not be used. Xen 3.3 runs on i386 PAE and amd64 hardware. There are no good reasons to run this version.
xenkernel41 provides Xen 4.1. It is no longer maintained by Xen, but as of 2016-12 received backported security patches. Xen 4.1 runs on i386 PAE and amd64 hardware. There are no good reasons to run this version.
Note that 3.1, 3.3 and 4.1 have been removed from pkgsrc-current, but are in 2016Q4. They will be removed from this HOWTO sometime after 2017Q1.
xenkernel42 provides Xen 4.2. It is no longer maintained by Xen, but as of 2016-12 received backported security patches. Xen 4.2 runs on i386 PAE and amd64 hardware. The only reason to run this is if you need to use xm instead of xl, or if you need to run on hardware that supports i386 but not amd64. (This might also be useful if you need an i386 dom0, if it turns out that an amd64 Xen kernel and an i386 dom0 is problematic.)
xenkernel45 provides Xen 4.5. As of 2016-12, security patches were released by Xen and applied to pkgsrc. Xen 4.5 runs on amd64 hardware only. While slightly old, 4.5 has been tested and run by others, so it is the conservative choice.
xenkernel46 provides Xen 4.6. It is new to pkgsrc as of 2016-05. As of 2016-12, security patches were released by Xen and applied to pkgsrc. Xen 4.6 runs on amd64 hardware only For new installations, 4.6 is probably the appropriate choice and it will likely soon be the standard approach. (If using Ubuntu guests, be sure to have the xentools46 from December, 2016).
Xen 4.7 (released 2016-06) and 4.8 (released 2016-12) are not yet in pkgsrc.
See also the Xen Security Advisory page.
Note that NetBSD support is called XEN3. It works with Xen 3 and Xen 4 because the hypercall interface has been stable.
Xen command program
Early Xen used a program called xm to manipulate the system from the dom0. Starting in 4.1, a replacement program with similar behavior called xl is provided, but it does not work well in 4.1. In 4.2, both xm and xl work fine. 4.4 is the last version that has xm.
You must make a global choice to use xm or xl, because it affects not only which command you use, but the command used by rc.d scripts (specifically xendomains) and which daemons should be run. The xentools packages provide xm for 3.1, 3.3 and 4.1 and xl for 4.2 and up.
In 4.2, you can choose to use xm by simply changing the ctl_command variable and setting xend=YES in rc.conf.
With xl, virtual devices are configured in parallel, which can cause problems if they are written assuming serial operation (e.g., updating firewall rules without explicit locking). There is now locking for the provided scripts, which works for normal casses (e.g, file-backed xbd, where a vnd must be allocated). But, as of 201612, it has not been adequately tested for a complex custom setup with a large number of interfaces.
The netbsd-6, netbsd-7, and -current branches are all reasonable choices, with more or less the same considerations for non-Xen use. Therefore, netbsd-7 is recommended as the stable version of the most recent release for production use. In addition, netbsd-7 and -current have a important scheduler fix (in November of 2015) affecting contention between dom0 and domUs; see https://releng.netbsd.org/cgi-bin/req-7.cgi?show=1040 for a description. For those wanting to learn Xen or without production stability concerns, netbsd-7 is still likely most appropriate, but -current is also a reasonable choice. (Xen runs ok on netbsd-5, but the xentools packages are likely difficult to build, and netbsd-5 is not supported.)
As of NetBSD 6, a NetBSD domU will support multiple vcpus. There is no SMP support for NetBSD as dom0. (The dom0 itself doesn't really need SMP for dom0 functions; the lack of support is really a problem when using a dom0 as a normal computer.)
Xen itself can run on i386 (Xen < 4.2) or amd64 hardware (all Xen versions). (Practically, almost any computer where one would want to run Xen today supports amd64.)
Xen, the dom0 system, and each domU system can be either i386 or amd64. When building a xenkernel package, one obtains an i386 Xen kernel on an i386 host, and an amd64 Xen kernel on an amd64 host. If the Xen kernel is i386, then the dom0 kernel and all domU kernels must be i386. With an amd64 Xen kernel, an amd64 dom0 kernel is known to work, and an i386 dom0 kernel should in theory work. An amd64 Xen/dom0 is known to support both i386 and amd64 domUs.
i386 dom0 and domU kernels must be PAE (except for an i386 Xen 3.1 kernel, where one can use non-PAE for dom0 and all domUs); PAE kernels are included in the NetBSD default build. (Note that emacs (at least) fails if run on i386 with PAE when built without, and vice versa, presumably due to bugs in the undump code.)
Because of the above, the standard approach is to use an amd64 Xen kernel and NetBSD/amd64 for the dom0. For domUs, NetBSD/i386 (with the PAE kernel) and NetBSD/amd64 are in widespread use, and there is little to no Xen-specific reason to prefer one over the other.
Note that to use an i386 dom0 with Xen 4.5 or higher, one must build (or obtain from pre-built packages) an amd64 Xen kernel and install that on the system. (One must also use a PAE i386 kernel, but this is also required with an i386 Xen kernel.). Almost no one in the NetBSD/Xen community does this, and the standard, well-tested, approach is to use an amd64 dom0.
A posting on xen-devel explained that PV system call overhead was higher on amd64, and thus there is some notion that i386 guests are faster. It goes on to caution that the total situation is complex and not entirely understood. On top of that caution, the post is about Linux, not NetBSD. TODO: Include link to benchmarks, if someone posts them.
Mostly, NetBSD as a dom0 or domU is quite stable. However, there are some open PRs indicating problems.
Note also that there are issues with sparse vnd(4) instances, but these are not about Xen -- they just are noticed with sparse vnd(4) instances in support of virtual disks in a dom0.
Therefore, this HOWTO recommends running xenkernel45 or xenkernel46, xl, the NetBSD 7 stable branch, and to use an amd64 kernel as the dom0. Either the i386PAE or amd64 version of NetBSD may be used as domUs.
Because bugs are fixed quite often, and because of Xen security advisories, it is good to stay up to date with NetBSD (tracking a stable branch), with the Xen kernel (tracking a Xen version via pkgsrc), and with the Xen tools. Specifically, NetBSD (-7 and -current) got an important fix affecting dom0/domU timesharing in November, 2015, and xentools46 got a fix to enable Ubuntu guests to boot in December, 2016.
Ideally, all versions of Xen in pkgsrc would build on all supported versions of NetBSD/amd64, to the point where this section would be silly. However, that has not always been the case. Besides aging code and aging compilers, qemu (included in xentools for HVM support) is difficult to build. Note that there is intentionally no data for 4.5+ up for i386, and often omits xentools info if the corresponding kernel fails.
The following table gives status, with the date last checked (generally on the most recent quarterly branch). The first code is "builds" if it builds ok, and "FAIL" for a failure to build. The second code/date only appears for xenkernel* and is "works" if it runs ok as a dom0 and can support a domU, and "FAIL" if it won't boot or run a domU.
xenkernel3 netbsd-6 i386 FAIL 201612 xenkernel33 netbsd-6 i386 FAIL 201612 xenkernel41 netbsd-6 i386 builds 201612 xenkernel42 netbsd-6 i386 builds 201612 xentools3 netbsd-6 i386 FAIL 201612 xentools33 netbsd-6 i386 FAIL 201612 xentools41 netbsd-6 i386 builds 201612 xentools42 netbsd-6 i386 FAIL 201612 xenkernel3 netbsd-7 i386 FAIL 201412 xenkernel33 netbsd-7 i386 FAIL 201412 xenkernel41 netbsd-7 i386 builds 201412 xenkernel42 netbsd-7 i386 builds 201412 xentools41 netbsd-7 i386 builds 201412 xentools42 netbsd-7 i386 ??FAIL 201412 xenkernel3 netbsd-6 amd64 FAIL 201612 xenkernel33 netbsd-6 amd64 FAIL 201612 xenkernel41 netbsd-6 amd64 builds 201612 works 201612 xenkernel42 netbsd-6 amd64 builds 201612 works 201612 xenkernel45 netbsd-6 amd64 builds 201612 xenkernel46 netbsd-6 amd64 builds 201612 xentools41 netbsd-6 amd64 builds 201612 xentools42 netbsd-6 amd64 builds 201612 xentools45 netbsd-6 amd64 builds 201612 xentools46 netbsd-6 amd64 FAIL 201612 xenkernel3 netbsd-7 amd64 builds 201612 xenkernel33 netbsd-7 amd64 builds 201612 xenkernel41 netbsd-7 amd64 builds 201612 xenkernel42 netbsd-7 amd64 builds 201612 xenkernel45 netbsd-7 amd64 builds 201612 xenkernel46 netbsd-7 amd64 builds 201612 xentools3 netbsd-7 amd64 builds 201612 xentools3-hvm netbsd-7 amd64 builds 201612 xentools33 netbsd-7 amd64 FAIL 201612 xentools41 netbsd-7 amd64 builds 201612 xentools42 netbsd-7 amd64 builds 201612 xentools45 netbsd-7 amd64 builds 201612 xentools46 netbsd-7 amd64 builds 201612
NetBSD as a dom0
NetBSD can be used as a dom0 and works very well. The following sections address installation, updating NetBSD, and updating Xen. Note that it doesn't make sense to talk about installing a dom0 OS without also installing Xen itself. We first address installing NetBSD, which is not yet a dom0, and then adding Xen, pivoting the NetBSD install to a dom0 install by just changing the kernel and boot configuration.
For experimenting with Xen, a machine with as little as 1G of RAM and 100G of disk can work. For running many domUs in productions, far more will be needed; e.g. 4-8G and 1T of disk is reasonable for a half-dozen domUs of 512M and 32G each. Basically, the RAM and disk have to be bigger than the sum of the RAM/disk needs of the dom0 and all the domUs.
Styles of dom0 operation
There are two basic ways to use Xen. The traditional method is for the dom0 to do absolutely nothing other than providing support to some number of domUs. Such a system was probably installed for the sole purpose of hosting domUs, and sits in a server room on a UPS.
The other way is to put Xen under a normal-usage computer, so that the dom0 is what the computer would have been without Xen, perhaps a desktop or laptop. Then, one can run domUs at will. Purists will deride this as less secure than the previous approach, and for a computer whose purpose is to run domUs, they are right. But Xen and a dom0 (without domUs) is not meaningfully less secure than the same things running without Xen. One can boot Xen or boot regular NetBSD alternately with little problems, simply refraining from starting the Xen daemons when not running Xen.
Note that NetBSD as dom0 does not support multiple CPUs. This will limit the performance of the Xen/dom0 workstation approach. In theory the only issue is that the "backend drivers" are not yet MPSAFE: http://mail-index.netbsd.org/netbsd-users/2014/08/29/msg015195.html
Installation of NetBSD
First, install NetBSD/amd64 just as you would if you were not using Xen. However, the partitioning approach is very important.
If you want to use RAIDframe for the dom0, there are no special issues for Xen. Typically one provides RAID storage for the dom0, and the domU systems are unaware of RAID. The 2nd-stage loader bootxx_* skips over a RAID1 header to find /boot from a file system within a RAID partition; this is no different when booting Xen.
There are 4 styles of providing backing storage for the virtual disks used by domUs: raw partitions, LVM, file-backed vnd(4), and SAN.
With raw partitions, one has a disklabel (or gpt) partition sized for each virtual disk to be used by the domU. (If you are able to predict how domU usage will evolve, please add an explanation to the HOWTO. Seriously, needs tend to change over time.)
One can use lvm(8) to create logical devices to use for domU disks. This is almost as efficient as raw disk partitions and more flexible. Hence raw disk partitions should typically not be used.
One can use files in the dom0 file system, typically created by dd'ing /dev/zero to create a specific size. This is somewhat less efficient, but very convenient, as one can cp the files for backup, or move them between dom0 hosts.
Finally, in theory one can place the files backing the domU disks in a SAN. (This is an invitation for someone who has done this to add a HOWTO page.)
Installation of Xen
In the dom0, install sysutils/xenkernel42 and sysutils/xentools42 from pkgsrc (or another matching pair). See the pkgsrc documentation for help with pkgsrc. Ensure that your packages are recent; the HOWTO does not contemplate old builds.
For Xen 3.1, support for HVM guests is in sysutils/xentool3-hvm. More recent versions have HVM support integrated in the main xentools package. It is entirely reasonable to run only PV guests.
Next you need to install the selected Xen kernel itself, which is installed by pkgsrc as "/usr/pkg/xen*-kernel/xen.gz". Copy it to /. For debugging, one may copy xen-debug.gz; this is conceptually similar to DIAGNOSTIC and DEBUG in NetBSD. xen-debug.gz is basically only useful with a serial console. Then, place a NetBSD XEN3_DOM0 kernel in /, copied from releasedir/amd64/binary/kernel/netbsd-XEN3_DOM0.gz of a NetBSD build. If using i386, use releasedir/i386/binary/kernel/netbsd-XEN3PAE_DOM0.gz. (If using Xen 3.1 and i386, you may use XEN3_DOM0 with the non-PAE Xen. But you should not use Xen 3.1.) Both xen and the NetBSD kernel may be (and typically are) left compressed.
In a dom0, kernfs is mandatory for xend to communicate with the kernel, so ensure that /kern is in fstab. (A standard NetBSD install should already mount /kern.)
Because you already installed NetBSD, you have a working boot setup with an MBR bootblock, either bootxx_ffsv1 or bootxx_ffsv2 at the beginning of your root file system, have /boot, and likely also /boot.cfg. (If not, fix before continuing!)
Add a line to to /boot.cfg to boot Xen. See boot.cfg(5) for an example. The basic line is
menu=Xen:load /netbsd-XEN3_DOM0.gz console=pc;multiboot /xen.gz dom0_mem=256M
which specifies that the dom0 should have 256M, leaving the rest to be allocated for domUs. To use a serial console, use
menu=Xen:load /netbsd-XEN3_DOM0.gz;multiboot /xen.gz dom0_mem=256M console=com1 com1=9600,8n1
which will use the first serial port for Xen (which counts starting from 1, unlike NetBSD which counts starting from 0), forcing speed/parity. Because the NetBSD command line lacks a "console=pc" argument, it will use the default "xencons" console device, which directs the console I/O through Xen to the same console device Xen itself uses (in this case, the serial port).
In an attempt to add performance, one can also add
to force only one vcpu to be provided (since NetBSD dom0 can't use more) and to pin that vcpu to a physical CPU. TODO: benchmark this.
Xen has many boot options, and other than dom0 memory and max_vcpus, they are generally not necessary.
As with non-Xen systems, you should have a line to boot /netbsd (a kernel that works without Xen). Consider a line to boot /netbsd.ok (a fallback version of the non-Xen kernel, updated manually when you are sure /netbsd is ok). Consider also a line to boot fallback versions of Xen and the dom0 kernel, but note that non-Xen NetBSD can be used to resolve Xen booting issues.
Probably you want a default=N line to choose Xen in the absence of intervention.
Now, reboot so that you are running a DOM0 kernel under Xen, rather than GENERIC without Xen.
Using grub (historic)
Before NetBSD's native bootloader could support Xen, the use of grub was recommended. If necessary, see the old grub information.
The HowTo on Installing into RAID-1 explains how to set up booting a dom0 with Xen using grub with NetBSD's RAIDframe. (This is obsolete with the use of NetBSD's native boot. Now, just create a system with RAID-1, and alter /boot.cfg as described above.)
Xen logs will be in /var/log/xen.
Now, you have a system that will boot Xen and the dom0 kernel, but not do anything else special. Make sure that you have rebooted into Xen. There will be no domUs, and none can be started because you still have to configure the dom0 daemons.
The daemons which should be run vary with Xen version and with whether one is using xm or xl. The Xen 3.1, 3.3 and 4.1 packages use xm. Xen 4.2 and up packages use xl. To use xm with 4.2, edit xendomains to use xm instead.
For 3.1 and 3.3, you should enable xend and xenbackendd:
For 4.1 and up, you should enable xencommons. Not enabling xencommons will result in a hang; it is necessary to hit ^C on the console to let the machine finish booting. If you are using xm (default in 4.1, or if you changed xendomains in 4.2), you should also enable xend:
xend=YES # only if using xm, and only installed <= 4.2 xencommons=YES
TODO: Recommend for/against xen-watchdog.
After you have configured the daemons and either started them (in the order given) or rebooted, use xm or xl to inspect Xen's boot messages, available resources, and running domains. An example with xl follows:
# xl dmesg [xen's boot info] # xl info [available memory, etc.] # xl list Name Id Mem(MB) CPU State Time(s) Console Domain-0 0 64 0 r---- 58.1
Issues with xencommons
xencommons starts xenstored, which stores data on behalf of dom0 and domUs. It does not currently work to stop and start xenstored. Certainly all domUs should be shutdown first, following the sort order of the rc.d scripts. However, the dom0 sets up state with xenstored, and is not notified when xenstored exits, leading to not recreating the state when the new xenstored starts. Until there's a mechanism to make this work, one should not expect to be able to restart xenstored (and thus xencommons). There is currently no reason to expect that this will get fixed any time soon.
No-longer needed advice about devices
The installation of NetBSD should already have created devices for xen (xencons, xenevt, xsd_kva), but if they are not present, create them:
cd /dev && sh MAKEDEV xen
anita (for testing NetBSD)
With the setup so far (assuming 4.2/xl), one should be able to run anita (see pkgsrc/misc/py-anita) to test NetBSD releases, by doing (as root, because anita must create a domU):
anita --vmm=xl test file:///usr/obj/i386/
Alternatively, one can use --vmm=xm to use xm-based domU creation instead (and must, on Xen <= 4.1). TODO: confirm that anita xl really works.
Xen-specific NetBSD issues
There are (at least) two additional things different about NetBSD as a dom0 kernel compared to hardware.
One is that the module ABI is different because some of the #defines change, so one must build modules for Xen. As of netbsd-7, the build system does this automatically. TODO: check this. (Before building Xen modules was added, it was awkward to use modules to the point where it was considered that it did not work.)
The other difference is that XEN3_DOM0 does not have exactly the same options as GENERIC. While it is debatable whether or not this is a bug, users should be aware of this and can simply add missing config items if desired.
Updating NetBSD in a dom0
This is just like updating NetBSD on bare hardware, assuming the new version supports the version of Xen you are running. Generally, one replaces the kernel and reboots, and then overlays userland binaries and adjusts /etc.
Note that one must update both the non-Xen kernel typically used for rescue purposes and the DOM0 kernel used with Xen.
Converting from grub to /boot
These instructions were [TODO: will be] used to convert a system from grub to /boot. The system was originally installed in February of 2006 with a RAID1 setup and grub to boot Xen 2, and has been updated over time. Before these commands, it was running NetBSD 6 i386, Xen 4.1 and grub, much like the message linked earlier in the grub section.
# Install MBR bootblocks on both disks. fdisk -i /dev/rwd0d fdisk -i /dev/rwd1d # Install NetBSD primary boot loader (/ is FFSv1) into RAID1 components. installboot -v /dev/rwd0d /usr/mdec/bootxx_ffsv1 installboot -v /dev/rwd1d /usr/mdec/bootxx_ffsv1 # Install secondary boot loader cp -p /usr/mdec/boot / # Create boot.cfg following earlier guidance: menu=Xen:load /netbsd-XEN3PAE_DOM0.gz console=pc;multiboot /xen.gz dom0_mem=256M menu=Xen.ok:load /netbsd-XEN3PAE_DOM0.ok.gz console=pc;multiboot /xen.ok.gz dom0_mem=256M menu=GENERIC:boot menu=GENERIC single-user:boot -s menu=GENERIC.ok:boot netbsd.ok menu=GENERIC.ok single-user:boot netbsd.ok -s menu=Drop to boot prompt:prompt default=1 timeout=30
TODO: actually do this and fix it if necessary.
Upgrading Xen versions
Minor version upgrades are trivial. Just rebuild/replace the xenkernel version and copy the new xen.gz to / (where /boot.cfg references it), and reboot.
Major version upgrades are conceptually not difficult, but can run into all the issues found when installing Xen. Assuming migration from 4.1 to 4.2, remove the xenkernel41 and xentools41 packages and install the xenkernel42 and xentools42 packages. Copy the 4.2 xen.gz to /.
Ensure that the contents of /etc/rc.d/xen* are correct. Specifically, they must match the package you just installed and not be left over from some previous installation.
Enable the correct set of daemons; see the configuring section above. (Upgrading from 3.x to 4.x without doing this will result in a hang.)
Ensure that the domU config files are valid for the new version. Specifically, for 4.x remove autorestart=True, and ensure that disks are specified with numbers as the second argument, as the examples above show, and not NetBSD device names.
Hardware known to work
Arguably, this section is misplaced, and there should be a page of hardware that runs NetBSD/amd64 well, with the mostly-well-founded assumption that NetBSD/xen runs fine on any modern hardware that NetBSD/amd64 runs well on. Until then, we give motherboard/CPU (and sometimes RAM) pairs/triples to aid those choosing a motherboard. Note that Xen systems usually do not run X, so a listing here does not imply that X works at all.
Supermicro X9SRL-F, Xeon E5-1650 v2, 96 GiB ECC Supermicro ??, Atom C2758 (8 core), 32 GiB ECC ASUS M5A78L-M/USB3 AM3+ microATX, AMD Piledriver X8 4000MHz, 16 GiB ECC
Intel D915GEV, Pentium4 CPU 3.40GHz, 4GB 533MHz Synchronous DDR2 INTEL DG33FB, "Intel(R) Core(TM)2 Duo CPU E6850 @ 3.00GHz" INTEL DG33FB, "Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz"
Running Xen under qemu
The astute reader will note that this section is somewhat twisted. However, it can be useful to run Xen under qemu either because the version of NetBSD as a dom0 does not run on the hardware in use, or to generate automated test cases involving Xen.
In 2015-01, the following combination was reported to mostly work:
host OS: NetBSD/amd64 6.1.4 qemu: 2.2.0 from pkgsrc Xen kernel: xenkernel42-4.2.5nb1 from pkgsrc dom0 kernel: NetBSD/amd64 6.1.5 Xen tools: xentools42-4.2.5 from pkgsrc
See PR 47720 for a problem with dom0 shutdown.
Unprivileged domains (domU)
This section describes general concepts about domUs. It does not address specific domU operating systems or how to install them. The config files for domUs are typically in /usr/pkg/etc/xen, and are typically named so that the file name, domU name and the domU's host name match.
The domU is provided with CPU and memory by Xen, configured by the dom0. The domU is provided with disk and network by the dom0, mediated by Xen, and configured in the dom0.
Entropy in domUs can be an issue; physical disks and network are on the dom0. NetBSD's /dev/random system works, but is often challenged.
There is no good order to present config files and the concepts surrounding what is being configured. We first show an example config file, and then in the various sections give details.
See (at least in xentools41) /usr/pkg/share/examples/xen/xmexample*, for a large number of well-commented examples, mostly for running GNU/Linux.
The following is an example minimal domain configuration file "/usr/pkg/etc/xen/foo". It is (with only a name change) an actual known working config file on Xen 4.1 (NetBSD 5 amd64 dom0 and NetBSD 5 i386 domU). The domU serves as a network file server.
# -*- mode: python; -*- kernel = "/netbsd-XEN3PAE_DOMU-i386-foo.gz" memory = 1024 vif = [ 'mac=aa:00:00:d1:00:09,bridge=bridge0' ] disk = [ 'file:/n0/xen/foo-wd0,0x0,w', 'file:/n0/xen/foo-wd1,0x1,w' ]
The domain will have the same name as the file. The kernel has the host/domU name in it, so that on the dom0 one can update the various domUs independently. The vif line causes an interface to be provided, with a specific mac address (do not reuse MAC addresses!), in bridge mode. Two disks are provided, and they are both writable; the bits are stored in files and Xen attaches them to a vnd(4) device in the dom0 on domain creation. The system treats xbd0 as the boot device without needing explicit configuration.
By default xm looks for domain config files in /usr/pkg/etc/xen. Note that "xm create" takes the name of a config file, while other commands take the name of a domain. To create the domain, connect to the console, create the domain while attaching the console, shutdown the domain, and see if it has finished stopping, do (or xl with Xen >= 4.2):
xm create foo xm console foo xm create -c foo xm shutdown foo xm list
Typing ^] will exit the console session. Shutting down a domain is equivalent to pushing the power button; a NetBSD domU will receive a power-press event and do a clean shutdown. Shutting down the dom0 will trigger controlled shutdowns of all configured domUs.
On a physical computer, the BIOS reads sector 0, and a chain of boot loaders finds and loads a kernel. Normally this comes from the root file system. With Xen domUs, the process is totally different. The normal path is for the domU kernel to be a file in the dom0's file system. At the request of the dom0, Xen loads that kernel into a new domU instance and starts execution. While domU kernels can be anyplace, reasonable places to store domU kernels on the dom0 are in / (so they are near the dom0 kernel), in /usr/pkg/etc/xen (near the config files), or in /u0/xen (where the vdisks are).
Note that loading the domU kernel from the dom0 implies that boot blocks, /boot, /boot.cfg, and so on are all ignored in the domU. See the VPS section near the end for discussion of alternate ways to obtain domU kernels.
CPU and memory
A domain is provided with some number of vcpus, less than the number of CPUs seen by the hypervisor. (For a dom0, this is controlled by the boot argument "dom0_max_vcpus=1".) For a domU, it is controlled from the config file by the "vcpus = N" directive.
A domain is provided with memory; this is controlled in the config file by "memory = N" (in megabytes). In the straightforward case, the sum of the the memory allocated to the dom0 and all domUs must be less than the available memory.
Xen also provides a "balloon" driver, which can be used to let domains use more memory temporarily. TODO: Explain better, and explain how well it works with NetBSD.
With the file/vnd style, typically one creates a directory, e.g. /u0/xen, on a disk large enough to hold virtual disks for all domUs. Then, for each domU disk, one writes zeros to a file that then serves to hold the virtual disk's bits; a suggested name is foo-xbd0 for the first virtual disk for the domU called foo. Writing zeros to the file serves two purposes. One is that preallocating the contents improves performance. The other is that vnd on sparse files has failed to work. TODO: give working/notworking NetBSD versions for sparse vnd and gnats reference. Note that the use of file/vnd for Xen is not really different than creating a file-backed virtual disk for some other purpose, except that xentools handles the vnconfig commands. To create an empty 4G virtual disk, simply do
dd if=/dev/zero of=foo-xbd0 bs=1m count=4096
Do not use qemu-img-xen, because this will create sparse file. There have been recent (2015) reports of sparse vnd(4) devices causing lockups, but there is apparently no PR.
With the lvm style, one creates logical devices. They are then used similarly to vnds. TODO: Add an example with lvm.
In domU config files, the disks are defined as a sequence of 3-tuples. The first element is "method:/path/to/disk". Common methods are "file:" for file-backed vnd. and "phy:" for something that is already a (TODO: character or block) device.
The second element is an artifact of how virtual disks are passed to Linux, and a source of confusion with NetBSD Xen usage. Linux domUs are given a device name to associate with the disk, and values like "hda1" or "sda1" are common. In a NetBSD domU, the first disk appears as xbd0, the second as xbd1, and so on. However, xm/xl demand a second argument. The name given is converted to a major/minor by calling stat(2) on the name in /dev and this is passed to the domU. In the general case, the dom0 and domU can be different operating systems, and it is an unwarranted assumption that they have consistent numbering in /dev, or even that the dom0 OS has a /dev. With NetBSD as both dom0 and domU, using values of 0x0 for the first disk and 0x1 for the second works fine and avoids this issue. For a GNU/Linux guest, one can create /dev/hda1 in /dev, or to pass 0x301 for /dev/hda1.
The third element is "w" for writable disks, and "r" for read-only disks.
Note that NetBSD by default creates only vnd. If you need more than 4 total virtual disks at a time, run e.g. "./MAKEDEV vnd4" in the dom0.
Note that NetBSD by default creates only xbd. If you need more virtual disks in a domU, run e.g. "./MAKEDEV xbd4" in the domU.
Xen provides virtual Ethernets, each of which connects the dom0 and a domU. For each virtual network, there is an interface "xvifN.M" in the dom0, and in domU index N, a matching interface xennetM (NetBSD name). The interfaces behave as if there is an Ethernet with two adapters connected. From this primitive, one can construct various configurations. We focus on two common and useful cases for which there are existing scripts: bridging and NAT.
With bridging (in the example above), the domU perceives itself to be on the same network as the dom0. For server virtualization, this is usually best. Bridging is accomplished by creating a bridge(4) device and adding the dom0's physical interface and the various xvifN.0 interfaces to the bridge. One specifies "bridge=bridge0" in the domU config file. The bridge must be set up already in the dom0; an example /etc/ifconfig.bridge0 is:
create up !brconfig bridge0 add wm0
With NAT, the domU perceives itself to be behind a NAT running on the dom0. This is often appropriate when running Xen on a workstation. TODO: NAT appears to be configured by "vif = [ '' ]".
The MAC address specified is the one used for the interface in the new domain. The interface in dom0 will use this address XOR'd with 00:00:00:01:00:00. Random MAC addresses are assigned if not given.
Modern x86 hardware has vast amounts of resources. However, many virtual servers can function just fine on far less. A system with 256M of RAM and a 4G disk can be a reasonable choice. Note that it is far easier to adjust virtual resources than physical ones. For memory, it's just a config file edit and a reboot. For disk, one can create a new file and vnconfig it (or lvm), and then dump/restore, just like updating physical disks, but without having to be there and without those pesky connectors.
Starting domains automatically
To start domains foo at bar at boot and shut them down cleanly on dom0 shutdown, in rc.conf add:
Note that earlier versions of the xentools41 xendomains rc.d script used xl, when one should use xm with 4.1.
Creating specific unprivileged domains (domU)
Creating domUs is almost entirely independent of operating system. We have already presented the basics of config files. Note that you must have already completed the dom0 setup so that "xl list" (or "xm list") works.
Creating an unprivileged NetBSD domain (domU)
See the earlier config file, and adjust memory. Decide on how much storage you will provide, and prepare it (file or lvm).
While the kernel will be obtained from the dom0 file system, the same file should be present in the domU as /netbsd so that tools like savecore(8) can work. (This is helpful but not necessary.)
The kernel must be specifically for Xen and for use as a domU. The i386 and amd64 provide the following kernels:
i386 XEN3_DOMU i386 XEN3PAE_DOMU amd64 XEN3_DOMU
Unless using Xen 3.1 (and you shouldn't) with i386-mode Xen, you must use the PAE version of the i386 kernel.
This will boot NetBSD, but this is not that useful if the disk is empty. One approach is to unpack sets onto the disk outside of xen (by mounting it, just as you would prepare a physical disk for a system you can't run the installer on).
A second approach is to run an INSTALL kernel, which has a miniroot and can load sets from the network. To do this, copy the INSTALL kernel to / and change the kernel line in the config file to:
kernel = "/home/bouyer/netbsd-INSTALL_XEN3_DOMU"
Then, start the domain as "xl create -c configname".
Alternatively, if you want to install NetBSD/Xen with a CDROM image, the following line should be used in the config file.
disk = [ 'phy:/dev/wd0e,0x1,w', 'phy:/dev/cd0a,0x2,r' ]
After booting the domain, the option to install via CDROM may be
selected. The CDROM device should be changed to
Once done installing, "halt -p" the new domain (don't reboot or halt, it would reload the INSTALL_XEN3_DOMU kernel even if you changed the config file), switch the config file back to the XEN3_DOMU kernel, and start the new domain again. Now it should be able to use "root on xbd0a" and you should have a, functional NetBSD domU.
TODO: check if this is still accurate.
When the new domain is booting you'll see some warnings about wscons
and the pseudo-terminals. These can be fixed by editing the files
/etc/wscons.conf. You must disable all terminals in
/etc/ttys, except console, like this:
console "/usr/libexec/getty Pc" vt100 on secure ttyE0 "/usr/libexec/getty Pc" vt220 off secure ttyE1 "/usr/libexec/getty Pc" vt220 off secure ttyE2 "/usr/libexec/getty Pc" vt220 off secure ttyE3 "/usr/libexec/getty Pc" vt220 off secure
Finally, all screens must be commented out from
It is also desirable to add
in rc.conf. This way, the domain will be properly shut down if
xm shutdown -R or
xm shutdown -H is used on the dom0.
It is not strictly necessary to have a kernel (as /netbsd) in the domU file system. However, various programs (e.g. netstat) will use that kernel to look up symbols to read from kernel virtual memory. If /netbsd is not the running kernel, those lookups will fail. (This is not really a Xen-specific issue, but because the domU kernel is obtained from the dom0, it is far more likely to be out of sync or missing with Xen.)
Creating an unprivileged Linux domain (domU)
Creating unprivileged Linux domains isn't much different from unprivileged NetBSD domains, but there are some details to know.
First, the second parameter passed to the disk declaration (the '0x1' in the example below)
disk = [ 'phy:/dev/wd0e,0x1,w' ]
does matter to Linux. It wants a Linux device number here (e.g. 0x300 for hda). Linux builds device numbers as: (major \<< 8 + minor). So, hda1 which has major 3 and minor 1 on a Linux system will have device number 0x301. Alternatively, devices names can be used (hda, hdb, ...) as xentools has a table to map these names to devices numbers. To export a partition to a Linux guest we can use:
disk = [ 'phy:/dev/wd0e,0x300,w' ] root = "/dev/hda1 ro"
and it will appear as /dev/hda on the Linux system, and be used as root partition.
To install the Linux system on the partition to be exported to the
guest domain, the following method can be used: install
sysutils/e2fsprogs from pkgsrc. Use mke2fs to format the partition
that will be the root partition of your Linux domain, and mount it.
Then copy the files from a working Linux system, make adjustments in
/etc (fstab, network config). It should also be possible to extract
binary packages such as .rpm or .deb directly to the mounted partition
using the appropriate tool, possibly running under NetBSD's Linux
emulation. Once the file system has been populated, umount it. If
desirable, the file system can be converted to ext3 using tune2fs -j.
It should now be possible to boot the Linux guest domain, using one of
the vmlinuz-*-xenU kernels available in the Xen binary distribution.
To get the Linux console right, you need to add:
extra = "xencons=tty1"
to your configuration since not all Linux distributions auto-attach a tty to the xen console.
Creating an unprivileged Solaris domain (domU)
See possibly outdated Solaris domU instructions.
PCI passthrough: Using PCI devices in guest domains
The dom0 can give other domains access to selected PCI devices. This can allow, for example, a non-privileged domain to have access to a physical network interface or disk controller. However, keep in mind that giving a domain access to a PCI device most likely will give the domain read/write access to the whole physical memory, as PCs don't have an IOMMU to restrict memory access to DMA-capable device. Also, it's not possible to export ISA devices to non-dom0 domains, which means that the primary VGA adapter can't be exported. A guest domain trying to access the VGA registers will panic.
If the dom0 is NetBSD, it has to be running Xen 3.1, as support has not been ported to later versions at this time.
For a PCI device to be exported to a domU, is has to be attached to the "pciback" driver in dom0. Devices passed to the dom0 via the pciback.hide boot parameter will attach to "pciback" instead of the usual driver. The list of devices is specified as "(bus:dev.func)", where bus and dev are 2-digit hexadecimal numbers, and func a single-digit number:
pciback devices should show up in the dom0's boot messages, and the
devices should be listed in the
PCI devices to be exported to a domU are listed in the "pci" array of the domU's config file, with the format "0000:bus:dev.func".
pci = [ '0000:00:06.0', '0000:00:0a.0' ]
In the domU an "xpci" device will show up, to which one or more pci buses will attach. Then the PCI drivers will attach to PCI buses as usual. Note that the default NetBSD DOMU kernels do not have "xpci" or any PCI drivers built in by default; you have to build your own kernel to use PCI devices in a domU. Here's a kernel config example; note that only the "xpci" lines are unusual.
include "arch/i386/conf/XEN3_DOMU" # Add support for PCI buses to the XEN3_DOMU kernel xpci* at xenbus ? pci* at xpci ? # PCI USB controllers uhci* at pci? dev ? function ? # Universal Host Controller (Intel) # USB bus support usb* at uhci? # USB Hubs uhub* at usb? uhub* at uhub? port ? configuration ? interface ? # USB Mass Storage umass* at uhub? port ? configuration ? interface ? wd* at umass? # SCSI controllers ahc* at pci? dev ? function ? # Adaptec 94x, aic78x0 SCSI # SCSI bus support (for both ahc and umass) scsibus* at scsi? # SCSI devices sd* at scsibus? target ? lun ? # SCSI disk drives cd* at scsibus? target ? lun ? # SCSI CD-ROM drives
NetBSD as a domU in a VPS
The bulk of the HOWTO is about using NetBSD as a dom0 on your own hardware. This section explains how to deal with Xen in a domU as a virtual private server where you do not control or have access to the dom0. This is not intended to be an exhaustive list of VPS providers; only a few are mentioned that specifically support NetBSD.
VPS operators provide varying degrees of access and mechanisms for configuration. The big issue is usually how one controls which kernel is booted, because the kernel is nominally in the dom0 file system (to which VPS users do not normally have access). A second issue is how to install NetBSD. A VPS user may want to compile a kernel for security updates, to run npf, run IPsec, or any other reason why someone would want to change their kernel.
One approach is to have an administrative interface to upload a kernel, or to select from a prepopulated list. Other approaches are pygrub (deprecated) and pvgrub, which are ways to have a bootloader obtain a kernel from the domU file system. This is closer to a regular physical computer, where someone who controls a machine can replace the kernel.
A second issue is multiple CPUs. With NetBSD 6, domUs support multiple vcpus, and it is typical for VPS providers to enable multiple CPUs for NetBSD domUs.
pygrub runs in the dom0 and looks into the domU file system. This implies that the domU must have a kernel in a file system in a format known to pygrub. As of 2014, pygrub seems to be of mostly historical interest.
pvgrub is a version of grub that uses PV operations instead of BIOS calls. It is booted from the dom0 as the domU kernel, and then reads /grub/menu.lst and loads a kernel from the domU file system.
Panix lets users use pvgrub. Panix reports that pvgrub works with FFsv2 with 16K/2K and 32K/4K block/frag sizes (and hence with defaults from "newfs -O 2"). See Panix's pvgrub page, which describes only Linux but should be updated to cover NetBSD :-).
It appears that grub's FFS code does not support all aspects of modern FFS, but there are also reports that FFSv2 works fine. At prgmr, typically one has an ext2 or FAT partition for the kernel with the intent that grub can understand it, which leads to /netbsd not being the actual kernel. One must remember to update the special boot partition.
See the Amazon EC2 page.
In standard kernels, npf is a module, and thus cannot be loaded in a DOMU kernel.
TODO: Explain how to compile npf into a custom kernel, answering (but note that the problem was caused by not booting the right kernel) this email to netbsd-users.
TODO items for improving NetBSD/xen
- Make the NetBSD dom0 kernel work with SMP.
- Test the Xen 4.5 packages adequately to be able to recommend them as the standard approach.
- Get PCI passthrough working on Xen 4.5
- Get pvgrub into pkgsrc, either via xentools or separately.
- Check/add support to pkgsrc grub2 for UFS2 and arbitrary fragsize/blocksize (UFS2 support may be present; the point is to make it so that with any UFS1/UFS2 file system setup that works with NetBSD grub will also work). See pkg/40258.
- Push patches upstream.
- Get UFS2 patches into pvgrub.
- Add support for PV ops to a version of /boot, and make it usable as a kernel in Xen, similar to pvgrub.
- Solve somehow the issue with modules for GENERIC not being loadable in a Xen dom0 or domU kernel.
This section contains links from elsewhere not yet integrated into the HOWTO, and other guides.