Annotation of wikisrc/ports/xen/howto.mdwn, revision 1.204

1.203     gdt         1: [[!meta title="Xen Status and HowTo"]]
1.144     maxv        2: 
1.157     maxv        3: Xen is a Type 1 hypervisor which supports running multiple guest operating
                      4: systems on a single physical machine. One uses the Xen kernel to control the
                      5: CPU, memory and console, a dom0 operating system which mediates access to
                      6: other hardware (e.g., disks, network, USB), and one or more domU operating
                      7: systems which operate in an unprivileged virtualized environment. IO requests
                      8: from the domU systems are forwarded by the Xen hypervisor to the dom0 to be
                      9: fulfilled.
1.12      gdt        10: 
1.203     gdt        11: This document provides status on what Xen things work on NetBSD
                     12: (upstream documentation might say something works if it works on some
                     13: particular Linux system).
                     15: This document is also a HOWTO that presumes a basic familiarity with
                     16: the Xen system architecture, with installing NetBSD on amd64 hardware,
                     17: and with installing software from pkgsrc.  See also the [Xen
1.182     gdt        18: website](
1.203     gdt        20: If this document says that something works, and you find that it does
                     21: not, it is best to ask on port-xen and if you are correct to file a
                     22: PR.
1.201     gdt        23: 
1.182     gdt        24: [[!toc]]
1.183     gdt        26: # Overview
1.182     gdt        27: 
                     28: The basic concept of Xen is that the hypervisor (xenkernel) runs on
                     29: the hardware, and runs a privileged domain ("dom0") that can access
1.204   ! gdt        30: disks/networking/etc.  One then runs additional unprivileged domains
1.182     gdt        31: (each a "domU"), presumably to do something useful.
                     33: This HOWTO addresses how to run a NetBSD dom0 (and hence also build
                     34: xen itself).  It also addresses how to run domUs in that environment,
                     35: and how to deal with having a domU in a Xen environment run by someone
                     36: else and/or not running NetBSD.
1.186     gdt        38: There are many choices one can make; the HOWTO recommends the standard
                     39: approach and limits discussion of alternatives in many cases.
1.182     gdt        41: ## Guest Styles
                     43: Xen supports different styles of guests.
1.149     maxv       44: 
                     45: [[!table data="""
1.203     gdt        46: Style of guest |description                            |NetBSD dom0?   |NetBSD domU?
                     47: PV             |Paravirtualization (hypercalls)        |yes            |yes
1.204   ! gdt        48: HVM            |unmodified guest - domU perceives HW   |can run guests |yes
        !            49: PVHVM          |Like HVM but also PV drivers           |can run guests |current only
        !            50: PVH            |Lightweight HVM (no qemu) and PV       |not yet        |current only
1.149     maxv       51: """]]
                     53: In Para-Virtualized (PV) mode, the guest OS does not attempt to access
                     54: hardware directly, but instead makes hypercalls to the hypervisor; PV
1.170     gdt        55: guests must be specifically coded for Xen.
1.182     gdt        56: See [PV](\)).
1.170     gdt        57: 
                     58: In HVM mode, no guest modification is required; however, hardware
                     59: support is required, such as VT-x on Intel CPUs and SVM on AMD CPUs.
1.199     gdt        60: The dom0 runs qemu to emulate hardware.  It is therefore non-sensical
                     61: to have an HVM dom0.
1.170     gdt        62: 
1.176     gdt        63: In PVHVM mode, the guest runs as HVM, but additionally can use PV
1.199     gdt        64: drivers for efficiency.  Therefore it is non-sensical for to have a
                     65: PVHVM dom0.  See [PV on HVM](
1.170     gdt        66: 
1.180     gdt        67: There have been two PVH modes: original PVH and PVHv2.  Original PVH
1.203     gdt        68: was based on PV mode and is no longer relevant at all.  Therefore
                     69: PVHv2 is abreviated PVH.  PVHv2 is basically lightweight HVM with PV
                     70: drivers.  A critical feature of it is that qemu is not needed; the
                     71: hypervisor can do the emulation that is required.  Thus, a dom0 can be
                     72: PVHv2.  The source code uses PVH and config files use pvh, but NB that
                     73: this refers to PVHv2.  See
                     74: [PVH(v2)](\)_Domu).
1.180     gdt        75: 
1.182     gdt        76: At system boot, the dom0 kernel is loaded as a module with Xen as the kernel.
1.12      gdt        77: The dom0 can start one or more domUs.  (Booting is explained in detail
                     78: in the dom0 section.)
1.182     gdt        80: ## CPU Architecture
                     82: Xen runs on x86_64 hardware (the NetBSD amd64 port).
                     84: There is a concept of Xen running on ARM, but there are no reports of this working with NetBSD.
                     86: The dom0 system should be amd64.  (Instructions for i386PAE dom0 have been removed from the HOWTO.)
1.1       mspo       87: 
1.204   ! gdt        88: The domU can be i386 PAE or amd64.
        !            89: i386 PAE at one point was considered as [faster]( than amd64.
        !            90: However, as of 2021 it is normal to use amd64 as the domU architecture, and use of i386 is dwindling.
1.150     maxv       91: 
1.182     gdt        92: ## Xen Versions
1.15      gdt        93: 
1.111     wiz        94: In NetBSD, Xen is provided in pkgsrc, via matching pairs of packages
1.15      gdt        95: xenkernel and xentools.  We will refer only to the kernel versions,
                     96: but note that both packages must be installed together and must have
                     97: matching versions.
1.145     maxv       99: Versions available in pkgsrc:
1.85      gdt       100: 
1.145     maxv      101: [[!table data="""
1.184     gdt       102: Xen Version    |Package Name   |Xen CPU Support        |EOL'ed By Upstream
                    103: 4.11           |xenkernel411   |x86_64                 |No
                    104: 4.13           |xenkernel413   |x86_64                 |No
1.145     maxv      105: """]]
1.113     gdt       106: 
1.96      gdt       107: See also the [Xen Security Advisory page](
1.204   ! gdt       109: Older Xen had a python-based management tool called xm; this has been
        !           110: replaced by xl.
1.166     gdt       111: 
1.183     gdt       112: ## NetBSD versions
                    114: Xen has been supported in NetBSD for a long time, at least since 2005.
                    115: Initially Xen was PV only.
1.204   ! gdt       117: NetBSD Xen has always supported PV, in both dom0 and domU; for a long
        !           118: time this was the only way.  NetBSD >=8 as a dom0 supports HVM mode in
        !           119: domUs (HVM as a dom0 does not make sense).
1.203     gdt       120: 
                    121: Support for PVHVM and PVH is available only in NetBSD-current; this is
                    122: currently somewhat experimental, although PVHVM appears reasonably
                    123: solid.
                    125: NetBSD up to and including NetBSD 9 as a dom0 cannot safely run SMP.
                    126: Even if one added "options MULTIPROCESSOR" and configured multiple
                    127: vcpus, the kernel is likely to crash because of drivers without
                    128: adequate locking.
                    130: NetBSD-current supports SMP in dom0, and XEN3_DOM0 includes "options
                    131: MULTIPROCESSOR".
                    133: NetBSD (since NetBSD 6), when run as a domU, can run SMP, using
                    134: multiple CPUs if provided.  The XEN3_DOMU kernel is built
                    135: with "options MULITPROCESSOR".
1.185     gdt       136: 
1.191     gdt       137: Note that while Xen 4.13 is current, the kernel support is still
                    138: called XEN3, because the hypercall interface has not changed
                    139: significantly.
1.185     gdt       140: 
1.187     gdt       141: # Creating a NetBSD dom0
1.15      gdt       142: 
1.191     gdt       143: In order to install a NetBSD as a dom0, one first installs a normal
                    144: NetBSD system, and then pivot the install to a dom0 install by
                    145: changing the kernel and boot configuration.
1.15      gdt       146: 
1.142     gdt       147: In 2018-05, trouble booting a dom0 was reported with 256M of RAM: with
                    148: 512M it worked reliably.  This does not make sense, but if you see
                    149: "not ELF" after Xen boots, try increasing dom0 RAM.
1.181     gdt       151: ## Installation of NetBSD
1.13      gdt       152: 
1.191     gdt       153: [Install NetBSD/amd64](/guide/inst/) just as you would if you were not
                    154: using Xen.  Therefore, use the most recent release, or a build from
                    155: the most recent stable branch.  Alternatively, use -current, being
                    156: mindful of all the usual caveats of lower stability of current, and
                    157: likely a bit more so.
1.1       mspo      158: 
1.181     gdt       159: ## Installation of Xen
1.1       mspo      160: 
1.191     gdt       161: ### Building Xen
1.187     gdt       163: Use the most recent version of Xen in pkgsrc, unless the DESCR says that it is not suitable.
1.186     gdt       164: Therefore, choose 4.13.
                    165: In the dom0, install xenkernel413 and xentools413 from pkgsrc.
1.155     maxv      166: 
1.186     gdt       167: Once this is done, copy the Xen kernel from where pkgsrc puts it to
                    168: where the boot process will be able to find it:
1.155     maxv      169: 
                    170: [[!template id=programlisting text="""
1.186     gdt       171: # cp -p /usr/pkg/xen413-kernel/xen.gz /
1.155     maxv      172: """]]
1.191     gdt       174: Then, place a NetBSD XEN3_DOM0 kernel in the `/` directory. Such
                    175: kernel can either be taken from a local release run, compiled
                    176: manually, or downloaded from the NetBSD FTP, for example at:
1.159     maxv      177: 
                    178: [[!template id=programlisting text="""
1.186     gdt       179:
1.159     maxv      180: """]]
1.20      gdt       181: 
1.191     gdt       182: ### Configuring booting
1.198     gdt       184: Read boot.cfg(8) carefully.  Add lines to /boot.cfg to boot Xen,
                    185: adjusting for your root filesystem:
1.20      gdt       186: 
1.157     maxv      187: [[!template id=filecontent name="/boot.cfg" text="""
1.198     gdt       188: menu=Xen:load /netbsd-XEN3_DOM0.gz root=wd0a console=pc;multiboot /xen.gz dom0_mem=512M
                    189: menu=Xen single user:load /netbsd-XEN3_DOM0.gz root=wd0a console=pc -s;multiboot /xen.gz dom0_mem=512M
1.152     maxv      190: """]]
1.20      gdt       191: 
1.159     maxv      192: This specifies that the dom0 should have 512MB of ram, leaving the rest
1.187     gdt       193: to be allocated for domUs.
1.191     gdt       195: NB: This says add, not replace, so that you will be able to more
                    196: easily boot a NetBSD kernel without Xen.  Once Xen boots ok, you may
                    197: want to set it as default.  It is highly likely that you will have
                    198: trouble at some point, and keeping an up-to-date GENERIC for use in
                    199: fixing problems is the standard prudent approach.
1.194     gdt       201: \todo Explain why rndseed is not set with Xen as part of the dom0
                    202: subconfiguration.
1.198     gdt       204: Note that you are likely to have to set root= because the boot device
                    205: from /boot is not passed via Xen to the dom0 kernel.  With one disk,
                    206: it will work, but e.g. plugging in USB disk to a machine with root on
                    207: wd0a causes boot to fail.
1.195     gdt       209: Beware that userconf statements must be attached to the dom0 load, and
                    210: may not be at top-level, because then they would try to configure the
                    211: hypervisor, if there is a way to pass them via multiboot .  It appears
                    212: that adding `userconf=pckbc` to `/boot.cfg` causes Xen to crash very
                    213: early with a heap overflow.
1.191     gdt       215: ### Console selection
                    217: See boot_console(8).  Understand that you should start from a place of
                    218: having console setup correct for booting GENERIC before trying to
                    219: configure Xen.
1.193     gdt       221: Generally for GENERIC, one sets the console in bootxx_ffsv1 or
                    222: equivalent, and this is passed on to /boot (where one typically does
                    223: not set the console).  This configuration of bootxx_ffsv1 should also
                    224: be in place for Xen systems, to allow seeing messages from /boot and
                    225: use of a keyboard to select a line from the menu.  And, one should
                    226: have a working boot path to GENERIC for rescue situations.
                    228: With GENERIC, the boot options are passed on to /netbsd, but there is
                    229: currently no mechanism to pass these via multiboot to the hypervisor.
                    230: Thus, in addition to configuring the console in the boot blocks, one
                    231: must also configure it for Xen.
                    233: By default, the hypervisor (Xen itself) will use some sort of vga
                    234: device as the console, much like GENERIC uses by default.  The vga
                    235: console is relinquished at the conclusion of hypervisor boot, before
1.194     gdt       236: the dom0 is started.  Xen when using a vga console does not process
                    237: console input.
1.77      gdt       238: 
1.193     gdt       239: The hypervisor can be configured to use a serial port console, e.g.
1.157     maxv      240: [[!template id=filecontent name="/boot.cfg" text="""
1.202     gson      241: menu=Xen:load /netbsd-XEN3_DOM0.gz console=com0;multiboot /xen.gz dom0_mem=512M console=com1 com1=9600,8n1
1.152     maxv      242: """]]
1.194     gdt       243: This example uses the first serial port (Xen counts from 1; this is
1.193     gdt       244: what NetBSD would call com0), and sets speed and parity.  (The dom0 is
                    245: then configured to use the same serial port in this example.)
1.194     gdt       247: With the hypervisor configured for a serial console, it can get input,
                    248: and there is a notion of passing this input to the dom0.  \todo
                    249: Explain why, if Xen has a serial console, the dom0 console is
                    250: typically also configured to open that same serial port, instead of
                    251: getting the passthrough input via the xen console.
1.193     gdt       253: One also configures the console for the dom0.  While one might expect
                    254: console=pc to be default, following behavior of GENERIC, a hasty read
                    255: of the code suggests there is no default and booting without a
                    256: selected console might lead to a panic.  Also, there is merit in
                    257: explicit configuration.  Therefore the standard approach is to place
                    258: console=pc as part of the load statement for the dom0 kernel, or
                    259: alternatively console=com0.
                    261: The NetBSD dom0 kernel will attach xencons(4) (the man page does not
                    262: exist), but this is not used as a console.  It is used to obtain the
                    263: messages from the hypervisor's console; run `xl dmesg` to see them.
1.191     gdt       264: 
                    265: ### Tuning
1.157     maxv      267: In an attempt to add performance, one can also add `dom0_max_vcpus=1 dom0_vcpus_pin`,
1.37      gdt       268: to force only one vcpu to be provided (since NetBSD dom0 can't use
1.158     maxv      269: more) and to pin that vcpu to a physical CPU. Xen has
1.178     gdt       270: [many boot options](,
1.111     wiz       271: and other than dom0 memory and max_vcpus, they are generally not
1.93      gdt       272: necessary.
1.191     gdt       273: \todo Revisit this advice with current.
                    274: \todo Explain if anyone has ever actually measured that this helps.
                    276: ### rc.conf
1.93      gdt       277: 
1.186     gdt       278: Ensure that the boot scripts installed in
                    279: `/usr/pkg/share/examples/rc.d` are in `/etc/rc.d`, either because you
                    280: have `PKG_RCD_SCRIPTS=yes`, or manually.  (This is not special to Xen,
                    281: but a normal part of pkgsrc usage.)
1.159     maxv      282: 
1.186     gdt       283: Set `xencommons=YES` in rc.conf:
1.31      gdt       284: 
1.157     maxv      285: [[!template id=filecontent name="/etc/rc.conf" text="""
1.152     maxv      286: xencommons=YES
                    287: """]]
1.31      gdt       288: 
1.187     gdt       289: \todo Recommend for/against xen-watchdog.
1.191     gdt       291: ### Testing
1.158     maxv      293: Now, reboot so that you are running a DOM0 kernel under Xen, rather
                    294: than GENERIC without Xen.
1.157     maxv      295: 
1.158     maxv      296: Once the reboot is done, use `xl` to inspect Xen's boot messages,
1.157     maxv      297: available resources, and running domains.  For example:
1.34      gdt       298: 
1.153     maxv      299: [[!template id=programlisting text="""
                    300: # xl dmesg
                    301: ... xen's boot info ...
                    302: # xl info
                    303: ... available memory, etc ...
                    304: # xl list
                    305: Name              Id  Mem(MB)  CPU  State  Time(s)  Console
                    306: Domain-0           0       64    0  r----     58.1
                    307: """]]
                    309: Xen logs will be in /var/log/xen.
1.33      gdt       310: 
1.88      gdt       311: ### Issues with xencommons
1.157     maxv      313: `xencommons` starts `xenstored`, which stores data on behalf of dom0 and
1.88      gdt       314: domUs.  It does not currently work to stop and start xenstored.
                    315: Certainly all domUs should be shutdown first, following the sort order
                    316: of the rc.d scripts.  However, the dom0 sets up state with xenstored,
                    317: and is not notified when xenstored exits, leading to not recreating
                    318: the state when the new xenstored starts.  Until there's a mechanism to
                    319: make this work, one should not expect to be able to restart xenstored
                    320: (and thus xencommons).  There is currently no reason to expect that
                    321: this will get fixed any time soon.
1.187     gdt       322: \todo Confirm if this is still true in 2020.
1.82      gdt       323: 
1.181     gdt       324: ## Xen-specific NetBSD issues
1.40      gdt       325: 
                    326: There are (at least) two additional things different about NetBSD as a
                    327: dom0 kernel compared to hardware.
1.191     gdt       329: One is that through NetBSD 9 the module ABI is different because some
                    330: of the #defines change, so there are separate sets of modules in
                    331: /stand.  In NetBSD-current, there is only one set of modules.
1.40      gdt       332: 
                    333: The other difference is that XEN3_DOM0 does not have exactly the same
                    334: options as GENERIC.  While it is debatable whether or not this is a
                    335: bug, users should be aware of this and can simply add missing config
                    336: items if desired.
1.187     gdt       338: Finally, there have been occasional reports of trouble with X11
                    339: servers in NetBSD as a dom0.
                    341: ## Updating Xen in a dom0
                    343: Basically, update the xenkernel and xentools packages and copy the new
                    344: Xen kernel into place, and reboot.  This procedure should be usable to
                    345: update to a new Xen release, but the reader is reminded that having a
                    346: non-Xen boot methods was recommended earlier.
1.181     gdt       348: ## Updating NetBSD in a dom0
1.15      gdt       349: 
                    350: This is just like updating NetBSD on bare hardware, assuming the new
                    351: version supports the version of Xen you are running.  Generally, one
                    352: replaces the kernel and reboots, and then overlays userland binaries
1.157     maxv      353: and adjusts `/etc`.
1.15      gdt       354: 
1.191     gdt       355: Note that one should update both the non-Xen kernel typically used for
                    356: rescue purposes, as well as the DOM0 kernel used with Xen.
1.15      gdt       357: 
1.187     gdt       358: ## anita (for testing NetBSD)
                    360: With a NetBSD dom0, even without any domUs, one should be able to run
                    361: anita (see pkgsrc/misc/py-anita) to test NetBSD releases, by doing (as
                    362: root, because anita must create a domU):
                    364: [[!template id=programlisting text="""
                    365: anita --vmm=xl test file:///usr/obj/i386/
                    366: """]]
                    368: # Unprivileged domains (domU)
1.28      gdt       369: 
                    370: This section describes general concepts about domUs.  It does not
1.33      gdt       371: address specific domU operating systems or how to install them.  The
1.157     maxv      372: config files for domUs are typically in `/usr/pkg/etc/xen`, and are
1.60      wiki      373: typically named so that the file name, domU name and the domU's host
1.33      gdt       374: name match.
1.111     wiz       376: The domU is provided with CPU and memory by Xen, configured by the
1.33      gdt       377: dom0.  The domU is provided with disk and network by the dom0,
                    378: mediated by Xen, and configured in the dom0.
                    380: Entropy in domUs can be an issue; physical disks and network are on
                    381: the dom0.  NetBSD's /dev/random system works, but is often challenged.
1.181     gdt       383: ## Config files
1.48      gdt       384: 
1.200     gdt       385: See /usr/pkg/share/examples/xen/xlexample* for a very small number of
                    386: examples for running GNU/Linux.
1.48      gdt       387: 
1.157     maxv      388: The following is an example minimal domain configuration file. The domU
                    389: serves as a network file server.
                    391: [[!template id=filecontent name="/usr/pkg/etc/xen/foo" text="""
                    392: name = "domU-id"
                    393: kernel = "/netbsd-XEN3PAE_DOMU-i386-foo.gz"
                    394: memory = 1024
                    395: vif = [ 'mac=aa:00:00:d1:00:09,bridge=bridge0' ]
                    396: disk = [ 'file:/n0/xen/foo-wd0,0x0,w',
                    397:          'file:/n0/xen/foo-wd1,0x1,w' ]
                    398: """]]
1.48      gdt       399: 
1.157     maxv      400: The domain will have name given in the `name` setting.  The kernel has the
1.48      gdt       401: host/domU name in it, so that on the dom0 one can update the various
1.157     maxv      402: domUs independently.  The `vif` line causes an interface to be provided,
1.48      gdt       403: with a specific mac address (do not reuse MAC addresses!), in bridge
                    404: mode.  Two disks are provided, and they are both writable; the bits
                    405: are stored in files and Xen attaches them to a vnd(4) device in the
1.111     wiz       406: dom0 on domain creation.  The system treats xbd0 as the boot device
1.48      gdt       407: without needing explicit configuration.
1.200     gdt       409: There is not type line; that implicitly defines a pv domU.  Otherwise,
                    410: one sets type to the lower-case version of the domU type in the table
1.201     gdt       411: above, e.g. `type = "hvm"`.
1.187     gdt       412: 
1.163     gson      413: By convention, domain config files are kept in `/usr/pkg/etc/xen`.  Note
1.157     maxv      414: that "xl create" takes the name of a config file, while other commands
                    415: take the name of a domain.
                    417: Examples of commands:
1.48      gdt       418: 
1.157     maxv      419: [[!template id=programlisting text="""
1.163     gson      420: xl create /usr/pkg/etc/xen/foo
                    421: xl console domU-id
                    422: xl create -c /usr/pkg/etc/xen/foo
                    423: xl shutdown domU-id
1.157     maxv      424: xl list
                    425: """]]
                    427: Typing `^]` will exit the console session.  Shutting down a domain is
1.48      gdt       428: equivalent to pushing the power button; a NetBSD domU will receive a
                    429: power-press event and do a clean shutdown.  Shutting down the dom0
                    430: will trigger controlled shutdowns of all configured domUs.
1.181     gdt       432: ## CPU and memory
1.33      gdt       433: 
1.169     plunky    434: A domain is provided with some number of vcpus, up to the number
1.157     maxv      435: of CPUs seen by the hypervisor. For a domU, it is controlled
1.48      gdt       436: from the config file by the "vcpus = N" directive.
                    438: A domain is provided with memory; this is controlled in the config
                    439: file by "memory = N" (in megabytes).  In the straightforward case, the
                    440: sum of the the memory allocated to the dom0 and all domUs must be less
1.33      gdt       441: than the available memory.
1.197     gdt       443: ## Balloon driver
                    445: Xen provides a `balloon` driver, which can be used to let domains use
                    446: more memory temporarily.
                    448: \todo Explain how to set up a aystem to use the balloon scheme in a
                    449: useful manner.
1.28      gdt       450: 
1.181     gdt       451: ## Virtual disks
1.28      gdt       452: 
1.158     maxv      453: In domU config files, the disks are defined as a sequence of 3-tuples:
                    455:  * The first element is "method:/path/to/disk". Common methods are
1.162     gutterid  456:    "file:" for a file-backed vnd, and "phy:" for something that is already
                    457:    a device, such as an LVM logical volume.
1.158     maxv      458: 
                    459:  * The second element is an artifact of how virtual disks are passed to
                    460:    Linux, and a source of confusion with NetBSD Xen usage.  Linux domUs
                    461:    are given a device name to associate with the disk, and values like
                    462:    "hda1" or "sda1" are common.  In a NetBSD domU, the first disk appears
                    463:    as xbd0, the second as xbd1, and so on.  However, xl demands a
                    464:    second argument.  The name given is converted to a major/minor by
                    465:    calling stat(2) on the name in /dev and this is passed to the domU.
                    466:    In the general case, the dom0 and domU can be different operating
                    467:    systems, and it is an unwarranted assumption that they have consistent
                    468:    numbering in /dev, or even that the dom0 OS has a /dev.  With NetBSD
                    469:    as both dom0 and domU, using values of 0x0 for the first disk and 0x1
                    470:    for the second works fine and avoids this issue.  For a GNU/Linux
                    471:    guest, one can create /dev/hda1 in /dev, or to pass 0x301 for
                    472:    /dev/hda1.
1.48      gdt       473: 
1.158     maxv      474:  * The third element is "w" for writable disks, and "r" for read-only
                    475:    disks.
                    477: Example:
                    478: [[!template id=filecontent name="/usr/pkg/etc/xen/foo" text="""
                    479: disk = [ 'file:/n0/xen/foo-wd0,0x0,w' ]
                    480: """]]
1.28      gdt       481: 
1.127     gdt       482: Note that NetBSD by default creates only vnd[0123].  If you need more
                    483: than 4 total virtual disks at a time, run e.g. "./MAKEDEV vnd4" in the
                    484: dom0.
1.187     gdt       486: ## Virtual Networking
1.28      gdt       487: 
1.111     wiz       488: Xen provides virtual Ethernets, each of which connects the dom0 and a
1.46      gdt       489: domU.  For each virtual network, there is an interface "xvifN.M" in
1.158     maxv      490: the dom0, and a matching interface xennetM (NetBSD name) in domU index N.
                    491: The interfaces behave as if there is an Ethernet with two
1.111     wiz       492: adapters connected.  From this primitive, one can construct various
1.46      gdt       493: configurations.  We focus on two common and useful cases for which
                    494: there are existing scripts: bridging and NAT.
1.28      gdt       495: 
1.48      gdt       496: With bridging (in the example above), the domU perceives itself to be
                    497: on the same network as the dom0.  For server virtualization, this is
                    498: usually best.  Bridging is accomplished by creating a bridge(4) device
                    499: and adding the dom0's physical interface and the various xvifN.0
                    500: interfaces to the bridge.  One specifies "bridge=bridge0" in the domU
                    501: config file.  The bridge must be set up already in the dom0; an
                    502: example /etc/ifconfig.bridge0 is:
1.46      gdt       503: 
1.157     maxv      504: [[!template id=filecontent name="/etc/ifconfig.bridge0" text="""
                    505: create
                    506: up
                    507: !brconfig bridge0 add wm0
                    508: """]]
1.28      gdt       509: 
                    510: With NAT, the domU perceives itself to be behind a NAT running on the
                    511: dom0.  This is often appropriate when running Xen on a workstation.
1.48      gdt       512: TODO: NAT appears to be configured by "vif = [ '' ]".
1.28      gdt       513: 
1.49      gdt       514: The MAC address specified is the one used for the interface in the new
1.53      gdt       515: domain.  The interface in dom0 will use this address XOR'd with
1.49      gdt       516: 00:00:00:01:00:00.  Random MAC addresses are assigned if not given.
1.187     gdt       518: ## Starting domains automatically
1.28      gdt       519: 
1.158     maxv      520: To start domains `domU-netbsd` and `domU-linux` at boot and shut them
                    521: down cleanly on dom0 shutdown, add the following in rc.conf:
1.28      gdt       522: 
1.158     maxv      523: [[!template id=filecontent name="/etc/rc.conf" text="""
                    524: xendomains="domU-netbsd domU-linux"
                    525: """]]
1.28      gdt       526: 
1.188     gdt       527: # domU setup for specific systems
1.14      gdt       528: 
                    529: Creating domUs is almost entirely independent of operating system.  We
1.188     gdt       530: have already presented the basics of config files in the previous system.
                    532: Of course, this section presumes that you have a working dom0.
1.14      gdt       533: 
1.187     gdt       534: ## Creating a NetBSD PV domU
1.49      gdt       536: See the earlier config file, and adjust memory.  Decide on how much
1.162     gutterid  537: storage you will provide, and prepare it (file or LVM).
1.49      gdt       538: 
1.111     wiz       539: While the kernel will be obtained from the dom0 file system, the same
1.49      gdt       540: file should be present in the domU as /netbsd so that tools like
                    541: savecore(8) can work.   (This is helpful but not necessary.)
1.188     gdt       543: The kernel must be specifically built for Xen, to use PV interfacesas
                    544: a domU.  NetBSD release builds provide the following kernels:
1.49      gdt       545: 
                    546:         i386 XEN3PAE_DOMU
1.95      gdt       547:         amd64 XEN3_DOMU
1.5       mspo      548: 
1.49      gdt       549: This will boot NetBSD, but this is not that useful if the disk is
1.188     gdt       550: empty.  One approach is to unpack sets onto the disk outside of Xen
1.49      gdt       551: (by mounting it, just as you would prepare a physical disk for a
                    552: system you can't run the installer on).
                    554: A second approach is to run an INSTALL kernel, which has a miniroot
                    555: and can load sets from the network.  To do this, copy the INSTALL
                    556: kernel to / and change the kernel line in the config file to:
1.5       mspo      557: 
1.49      gdt       558:         kernel = "/home/bouyer/netbsd-INSTALL_XEN3_DOMU"
1.5       mspo      559: 
1.163     gson      560: Then, start the domain as "xl create -c configfile".
1.1       mspo      561: 
1.49      gdt       562: Alternatively, if you want to install NetBSD/Xen with a CDROM image, the following
                    563: line should be used in the config file.
1.1       mspo      564: 
1.3       mspo      565:     disk = [ 'phy:/dev/wd0e,0x1,w', 'phy:/dev/cd0a,0x2,r' ]
1.1       mspo      566: 
                    567: After booting the domain, the option to install via CDROM may be
1.49      gdt       568: selected.  The CDROM device should be changed to `xbd1d`.
1.1       mspo      569: 
1.188     gdt       570: Once done installing, "halt -p" the new domain (don't reboot or halt:
1.49      gdt       571: it would reload the INSTALL_XEN3_DOMU kernel even if you changed the
                    572: config file), switch the config file back to the XEN3_DOMU kernel,
                    573: and start the new domain again. Now it should be able to use "root on
1.188     gdt       574: xbd0a" and you should have a functional NetBSD domU.
1.1       mspo      575: 
1.49      gdt       576: TODO: check if this is still accurate.
1.1       mspo      577: When the new domain is booting you'll see some warnings about *wscons*
                    578: and the pseudo-terminals. These can be fixed by editing the files
1.5       mspo      579: `/etc/ttys` and `/etc/wscons.conf`. You must disable all terminals in
                    580: `/etc/ttys`, except *console*, like this:
1.1       mspo      581: 
1.3       mspo      582:     console "/usr/libexec/getty Pc"         vt100   on secure
                    583:     ttyE0   "/usr/libexec/getty Pc"         vt220   off secure
                    584:     ttyE1   "/usr/libexec/getty Pc"         vt220   off secure
                    585:     ttyE2   "/usr/libexec/getty Pc"         vt220   off secure
                    586:     ttyE3   "/usr/libexec/getty Pc"         vt220   off secure
1.1       mspo      587: 
1.5       mspo      588: Finally, all screens must be commented out from `/etc/wscons.conf`.
1.1       mspo      589: 
1.192     gdt       590: One should also run `powerd` in a domU, but this should not need
                    591: configuring.  With powerd, the domain will run a controlled shutdown
                    592: if `xl shutdown -R` or `xl shutdown -H` is used on the dom0, via
                    593: receiving a synthetic `power button pressed` signal.  In 9 and
                    594: current, `powerd` is run by default under Xen kernels (or if ACPI is
                    595: present), and it can be added to rc.conf if not.
1.1       mspo      596: 
1.92      gdt       597: It is not strictly necessary to have a kernel (as /netbsd) in the domU
1.111     wiz       598: file system.  However, various programs (e.g. netstat) will use that
1.92      gdt       599: kernel to look up symbols to read from kernel virtual memory.  If
                    600: /netbsd is not the running kernel, those lookups will fail.  (This is
                    601: not really a Xen-specific issue, but because the domU kernel is
                    602: obtained from the dom0, it is far more likely to be out of sync or
                    603: missing with Xen.)
1.187     gdt       605: Note that NetBSD by default creates only xbd[0123].  If you need more
                    606: virtual disks in a domU, run e.g. "./MAKEDEV xbd4" in the domU.
1.196     gdt       608: ## Creating a Linux PV domU
1.1       mspo      609: 
                    610: Creating unprivileged Linux domains isn't much different from
                    611: unprivileged NetBSD domains, but there are some details to know.
                    613: First, the second parameter passed to the disk declaration (the '0x1' in
                    614: the example below)
1.3       mspo      616:     disk = [ 'phy:/dev/wd0e,0x1,w' ]
1.1       mspo      617: 
                    618: does matter to Linux. It wants a Linux device number here (e.g. 0x300
1.49      gdt       619: for hda).  Linux builds device numbers as: (major \<\< 8 + minor).
                    620: So, hda1 which has major 3 and minor 1 on a Linux system will have
                    621: device number 0x301.  Alternatively, devices names can be used (hda,
                    622: hdb, ...)  as xentools has a table to map these names to devices
                    623: numbers.  To export a partition to a Linux guest we can use:
1.1       mspo      624: 
1.49      gdt       625:         disk = [ 'phy:/dev/wd0e,0x300,w' ]
                    626:         root = "/dev/hda1 ro"
1.1       mspo      627: 
                    628: and it will appear as /dev/hda on the Linux system, and be used as root
                    629: partition.
1.49      gdt       631: To install the Linux system on the partition to be exported to the
                    632: guest domain, the following method can be used: install
                    633: sysutils/e2fsprogs from pkgsrc.  Use mke2fs to format the partition
                    634: that will be the root partition of your Linux domain, and mount it.
                    635: Then copy the files from a working Linux system, make adjustments in
                    636: `/etc` (fstab, network config).  It should also be possible to extract
                    637: binary packages such as .rpm or .deb directly to the mounted partition
                    638: using the appropriate tool, possibly running under NetBSD's Linux
1.111     wiz       639: emulation.  Once the file system has been populated, umount it.  If
                    640: desirable, the file system can be converted to ext3 using tune2fs -j.
1.49      gdt       641: It should now be possible to boot the Linux guest domain, using one of
                    642: the vmlinuz-\*-xenU kernels available in the Xen binary distribution.
1.1       mspo      643: 
1.111     wiz       644: To get the Linux console right, you need to add:
1.1       mspo      645: 
1.3       mspo      646:     extra = "xencons=tty1"
1.1       mspo      647: 
1.111     wiz       648: to your configuration since not all Linux distributions auto-attach a
1.1       mspo      649: tty to the xen console.
1.180     gdt       651: ## Creating a NetBSD HVM domU
1.196     gdt       653: Use type='hvm', probably.  Use a GENERIC kernel within the disk image.
1.180     gdt       654: 
                    655: ## Creating a NetBSD PVH domU
1.196     gdt       657: This only works with a current kernel in the domU.
1.180     gdt       658: 
1.196     gdt       659: Use type='pvh'.  Probably, use a GENERIC kernel within the disk image,
                    660: which in current has PV support.
                    662: \todo Verify.
                    664: \todo Verify if one can have current PVH domU on a 9 dom0.
1.180     gdt       665: 
1.187     gdt       666: ## Creating a Solaris domU
1.1       mspo      667: 
1.50      gdt       668: See possibly outdated
                    669: [Solaris domU instructions](/ports/xen/howto-solaris/).
1.5       mspo      670: 
1.187     gdt       671: ## PCI passthrough: Using PCI devices in guest domains
1.1       mspo      672: 
1.180     gdt       673: NB: PCI passthrough only works on some Xen versions and as of 2020 it
1.196     gdt       674: is not clear that it works on any version in pkgsrc.  \todo Reports
1.180     gdt       675: confirming or denying this notion should be sent to port-xen@.
1.53      gdt       677: The dom0 can give other domains access to selected PCI
1.52      gdt       678: devices. This can allow, for example, a non-privileged domain to have
                    679: access to a physical network interface or disk controller.  However,
                    680: keep in mind that giving a domain access to a PCI device most likely
                    681: will give the domain read/write access to the whole physical memory,
                    682: as PCs don't have an IOMMU to restrict memory access to DMA-capable
1.53      gdt       683: device.  Also, it's not possible to export ISA devices to non-dom0
1.52      gdt       684: domains, which means that the primary VGA adapter can't be exported.
                    685: A guest domain trying to access the VGA registers will panic.
1.53      gdt       687: If the dom0 is NetBSD, it has to be running Xen 3.1, as support has
1.52      gdt       688: not been ported to later versions at this time.
                    690: For a PCI device to be exported to a domU, is has to be attached to
                    691: the "pciback" driver in dom0.  Devices passed to the dom0 via the
                    692: pciback.hide boot parameter will attach to "pciback" instead of the
                    693: usual driver.  The list of devices is specified as "(bus:dev.func)",
1.5       mspo      694: where bus and dev are 2-digit hexadecimal numbers, and func a
                    695: single-digit number:
1.1       mspo      696: 
1.52      gdt       697:         pciback.hide=(00:0a.0)(00:06.0)
1.1       mspo      698: 
1.52      gdt       699: pciback devices should show up in the dom0's boot messages, and the
1.5       mspo      700: devices should be listed in the `/kern/xen/pci` directory.
1.1       mspo      701: 
1.52      gdt       702: PCI devices to be exported to a domU are listed in the "pci" array of
                    703: the domU's config file, with the format "0000:bus:dev.func".
1.1       mspo      704: 
1.52      gdt       705:         pci = [ '0000:00:06.0', '0000:00:0a.0' ]
1.1       mspo      706: 
1.52      gdt       707: In the domU an "xpci" device will show up, to which one or more pci
1.111     wiz       708: buses will attach.  Then the PCI drivers will attach to PCI buses as
1.52      gdt       709: usual.  Note that the default NetBSD DOMU kernels do not have "xpci"
                    710: or any PCI drivers built in by default; you have to build your own
                    711: kernel to use PCI devices in a domU.  Here's a kernel config example;
                    712: note that only the "xpci" lines are unusual.
                    714:         include         "arch/i386/conf/XEN3_DOMU"
1.111     wiz       716:         # Add support for PCI buses to the XEN3_DOMU kernel
1.52      gdt       717:         xpci* at xenbus ?
                    718:         pci* at xpci ?
                    720:         # PCI USB controllers
                    721:         uhci*   at pci? dev ? function ?        # Universal Host Controller (Intel)
                    723:         # USB bus support
                    724:         usb*    at uhci?
                    726:         # USB Hubs
                    727:         uhub*   at usb?
                    728:         uhub*   at uhub? port ? configuration ? interface ?
                    730:         # USB Mass Storage
                    731:         umass*  at uhub? port ? configuration ? interface ?
                    732:         wd*     at umass?
                    733:         # SCSI controllers
                    734:         ahc*    at pci? dev ? function ?        # Adaptec [23]94x, aic78x0 SCSI
                    736:         # SCSI bus support (for both ahc and umass)
                    737:         scsibus* at scsi?
                    739:         # SCSI devices
                    740:         sd*     at scsibus? target ? lun ?      # SCSI disk drives
                    741:         cd*     at scsibus? target ? lun ?      # SCSI CD-ROM drives
1.1       mspo      742: 
1.189     gdt       744: # Miscellaneous Information
1.173     gdt       745: 
1.189     gdt       746: ## Nesting under Linux KVM
                    748: It is possible to run a Xen and a NetBSD dom0 under Linux KVM.  One
                    749: can enable virtio in the dom0 for greater speed.
                    751: ## Other nesting
                    753: In theory, any full emulation should be able to run Xen and a NetBSD
                    754: dom0.  The HOWTO does not currently have information about Xen XVM
                    755: mode, nvmm, qemu, Virtualbox, etc.
                    757: ## NetBSD 5 as domU
1.173     gdt       758: 
1.176     gdt       759: [NetBSD 5 is known to panic.](
1.175     gdt       760: (However, NetBSD 5 systems should be updated to a supported version.)
1.173     gdt       761: 
                    762: # NetBSD as a domU in a VPS
1.28      gdt       763: 
                    764: The bulk of the HOWTO is about using NetBSD as a dom0 on your own
                    765: hardware.  This section explains how to deal with Xen in a domU as a
                    766: virtual private server where you do not control or have access to the
1.70      gdt       767: dom0.  This is not intended to be an exhaustive list of VPS providers;
                    768: only a few are mentioned that specifically support NetBSD.
1.28      gdt       769: 
1.52      gdt       770: VPS operators provide varying degrees of access and mechanisms for
                    771: configuration.  The big issue is usually how one controls which kernel
1.111     wiz       772: is booted, because the kernel is nominally in the dom0 file system (to
                    773: which VPS users do not normally have access).  A second issue is how
1.70      gdt       774: to install NetBSD.
1.52      gdt       775: A VPS user may want to compile a kernel for security updates, to run
                    776: npf, run IPsec, or any other reason why someone would want to change
                    777: their kernel.
1.111     wiz       779: One approach is to have an administrative interface to upload a kernel,
1.68      gdt       780: or to select from a prepopulated list.  Other approaches are pygrub
1.59      gdt       781: (deprecated) and pvgrub, which are ways to have a bootloader obtain a
1.111     wiz       782: kernel from the domU file system.  This is closer to a regular physical
1.59      gdt       783: computer, where someone who controls a machine can replace the kernel.
1.52      gdt       784: 
1.74      gdt       785: A second issue is multiple CPUs.  With NetBSD 6, domUs support
                    786: multiple vcpus, and it is typical for VPS providers to enable multiple
                    787: CPUs for NetBSD domUs.
1.180     gdt       789: ## Complexities due to Xen changes
                    791: Xen has many security advisories and people running Xen systems make
                    792: different choices.
                    794: ### stub domains
                    796: Some (Linux only?) dom0 systems use something called "stub domains" to
                    797: isolate qemu from the dom0 system, as a security and reliabilty
                    798: mechanism when running HVM domUs.  Somehow, NetBSD's GENERIC kernel
                    799: ends up using PIO for disks rather than DMA.  Of course, all of this
                    800: is emulated, but emulated PIO is unusably slow.  This problem is not
                    801: currently understood.
                    803: ### Grant tables
                    805: There are multiple versions of using grant tables, and some security
                    806: advisories have suggested disabling some versions.  Some versions of
                    807: NetBSD apparently only use specific versions and this can lead to
                    808: "NetBSD current doesn't run on hosting provider X" situations.
                    810: \todo Explain better.
1.181     gdt       812: ## Boot methods
                    814: ### pvgrub
1.59      gdt       815: 
                    816: pvgrub is a version of grub that uses PV operations instead of BIOS
                    817: calls.  It is booted from the dom0 as the domU kernel, and then reads
1.111     wiz       818: /grub/menu.lst and loads a kernel from the domU file system.
1.59      gdt       819: 
1.70      gdt       820: [Panix]( lets users use pvgrub.  Panix reports
1.71      gdt       821: that pvgrub works with FFsv2 with 16K/2K and 32K/4K block/frag sizes
                    822: (and hence with defaults from "newfs -O 2").  See [Panix's pvgrub
1.70      gdt       823: page](, which describes only
1.74      gdt       824: Linux but should be updated to cover NetBSD :-).
1.70      gdt       825: 
                    826: []( also lets users with pvgrub to boot
                    827: their own kernel.  See then [ NetBSD
1.74      gdt       828: HOWTO](
                    829: (which is in need of updating).
1.59      gdt       830: 
1.70      gdt       831: It appears that [grub's FFS
                    832: code](
                    833: does not support all aspects of modern FFS, but there are also reports
1.72      gdt       834: that FFSv2 works fine.  At prgmr, typically one has an ext2 or FAT
1.70      gdt       835: partition for the kernel with the intent that grub can understand it,
                    836: which leads to /netbsd not being the actual kernel.  One must remember
1.111     wiz       837: to update the special boot partition.
1.59      gdt       838: 
1.181     gdt       839: ### pygrub
1.168     wiki      840: 
                    841: pygrub runs in the dom0 and looks into the domU file system.  This
                    842: implies that the domU must have a kernel in a file system in a format
                    843: known to pygrub.
                    845: pygrub doesn't seem to work to load Linux images under NetBSD dom0,
                    846: and is inherently less secure than pvgrub due to running inside dom0. For both these
                    847: reasons, pygrub should not be used, and is only still present so that
                    848: historical DomU images using it still work.
                    850: As of 2014, pygrub seems to be of mostly historical
                    851: interest. New DomUs should use pvgrub.
1.181     gdt       853: ## Specific Providers
                    855: ### Amazon
1.59      gdt       856: 
1.143     wiki      857: See the [Amazon EC2 page](/amazon_ec2/).

CVSweb for NetBSD wikisrc <> software: FreeBSD-CVSweb