Annotation of wikisrc/ports/xen/howto.mdwn, revision 1.140

1.5       mspo        1: Introduction
1.13      gdt         2: ============
1.1       mspo        3: 
                      4: [![[Xen
1.140   ! khorben     5: screenshot]](https://www.netbsd.org/gallery/in-Action/hubertf-xens.png)](https://www.netbsd.org/gallery/in-Action/hubertf-xen.png)
1.1       mspo        6: 
1.58      gdt         7: Xen is a hypervisor (or virtual machine monitor) for x86 hardware
1.12      gdt         8: (i686-class or higher), which supports running multiple guest
1.58      gdt         9: operating systems on a single physical machine.  Xen is a Type 1 or
                     10: bare-metal hypervisor; one uses the Xen kernel to control the CPU,
                     11: memory and console, a dom0 operating system which mediates access to
                     12: other hardware (e.g., disks, network, USB), and one or more domU
                     13: operating systems which operate in an unprivileged virtualized
                     14: environment.  IO requests from the domU systems are forwarded by the
                     15: hypervisor (Xen) to the dom0 to be fulfilled.
1.12      gdt        16: 
                     17: Xen supports two styles of guests.  The original is Para-Virtualized
                     18: (PV) which means that the guest OS does not attempt to access hardware
                     19: directly, but instead makes hypercalls to the hypervisor.  This is
                     20: analogous to a user-space program making system calls.  (The dom0
                     21: operating system uses PV calls for some functions, such as updating
                     22: memory mapping page tables, but has direct hardware access for disk
                     23: and network.)   PV guests must be specifically coded for Xen.
                     24: 
                     25: The more recent style is HVM, which means that the guest does not have
                     26: code for Xen and need not be aware that it is running under Xen.
                     27: Attempts to access hardware registers are trapped and emulated.  This
                     28: style is less efficient but can run unmodified guests.
                     29: 
1.123     gdt        30: Generally any machine that runs NetBSD/amd64 will work with Xen and PV
                     31: guests.  In theory i386 computers (without x86_64/amd64 support) can
                     32: be used for Xen <= 4.2, but we have no recent reports of this working
                     33: (this is a hint).  For HVM guests, hardware support is needed, but it
                     34: is common on recent machines.  For Intel CPUs, one needs the VT-x
                     35: extension, shown in "cpuctl identify 0" as VMX.  For AMD CPUs, one
                     36: needs the AMD-V extensions, shown in "cpuctl identify 0" as SVM.
                     37: There are further features for IOMMU virtualization, Intel's VT-d and
                     38: AMD's AMD-Vi.  TODO: Explain whether Xen on NetBSD makes use of these
                     39: features.  TODO: Review by someone who really understands this.
1.19      gdt        40: 
1.123     gdt        41: Note that a FreeBSD dom0 requires VT-x and VT-d (or equivalent); this
                     42: is because the FreeBSD dom0 does not run in PV mode.
1.118     gdt        43: 
1.27      jnemeth    44: At boot, the dom0 kernel is loaded as a module with Xen as the kernel.
1.12      gdt        45: The dom0 can start one or more domUs.  (Booting is explained in detail
                     46: in the dom0 section.)
                     47: 
                     48: NetBSD supports Xen in that it can serve as dom0, be used as a domU,
                     49: and that Xen kernels and tools are available in pkgsrc.  This HOWTO
                     50: attempts to address both the case of running a NetBSD dom0 on hardware
1.24      gdt        51: and running domUs under it (NetBSD and other), and also running NetBSD
                     52: as a domU in a VPS.
1.12      gdt        53: 
1.123     gdt        54: Xen 3.1 in pkgsrc supports "PCI passthrough", which means that
1.20      gdt        55: specific PCI devices can be made available to a specific domU instead
                     56: of the dom0.  This can be useful to let a domU run X11, or access some
                     57: network interface or other peripheral.
                     58: 
1.111     wiz        59: NetBSD 6 and earlier supported Xen 2; support was removed from NetBSD
1.104     gdt        60: 7.  Xen 2 has been removed from pkgsrc.
1.54      gdt        61: 
1.12      gdt        62: Prerequisites
1.13      gdt        63: -------------
1.12      gdt        64: 
                     65: Installing NetBSD/Xen is not extremely difficult, but it is more
                     66: complex than a normal installation of NetBSD.
1.15      gdt        67: In general, this HOWTO is occasionally overly restrictive about how
                     68: things must be done, guiding the reader to stay on the established
                     69: path when there are no known good reasons to stray.
1.12      gdt        70: 
                     71: This HOWTO presumes a basic familiarity with the Xen system
1.123     gdt        72: architecture, with installing NetBSD on i386/amd64 hardware, and with
                     73: installing software from pkgsrc.  See also the [Xen
                     74: website](http://www.xenproject.org/).
1.1       mspo       75: 
1.15      gdt        76: Versions of Xen and NetBSD
                     77: ==========================
                     78: 
1.27      jnemeth    79: Most of the installation concepts and instructions are independent
                     80: of Xen version and NetBSD version.  This section gives advice on
                     81: which version to choose.  Versions not in pkgsrc and older unsupported
                     82: versions of NetBSD are intentionally ignored.
1.15      gdt        83: 
1.133     gdt        84: The term "amd64" is used to refer to both the NetBSD port and to the
                     85: hardware architecture on which it runs.  (Such hardware is made by
                     86: both Intel and AMD, and in 2016 a normal PC has this CPU
                     87: architecture.)
                     88: 
1.15      gdt        89: Xen
                     90: ---
                     91: 
1.111     wiz        92: In NetBSD, Xen is provided in pkgsrc, via matching pairs of packages
1.15      gdt        93: xenkernel and xentools.  We will refer only to the kernel versions,
                     94: but note that both packages must be installed together and must have
                     95: matching versions.
                     96: 
1.133     gdt        97: xenkernel3 provides Xen 3.1.  It is no longer maintained by Xen, and
                     98: the last applied security patch was in 2011. Thus, it should not be
                     99: used.  It supports PCI passthrough, which is why people use it anyway.
                    100: Xen 3.1 runs on i386 (both non-PAE and PAE) and amd64 hardware.
1.104     gdt       101: 
                    102: xenkernel33 provides Xen 3.3.  It is no longer maintained by Xen, and
                    103: the last applied security patch was in 2012.  Thus, it should not be
1.133     gdt       104: used.  Xen 3.3 runs on i386 PAE and amd64 hardware.  There are no good
1.104     gdt       105: reasons to run this version.
                    106: 
                    107: xenkernel41 provides Xen 4.1.  It is no longer maintained by Xen, but
1.133     gdt       108: as of 2016-12 received backported security patches.  Xen 4.1 runs on
                    109: i386 PAE and amd64 hardware.  There are no good reasons to run this
1.104     gdt       110: version.
1.102     gdt       111: 
1.137     gdt       112: Note that 3.1, 3.3 and 4.1 have been removed from pkgsrc-current, but
                    113: are in 2016Q4.  They will be removed from this HOWTO sometime after
1.138     wiz       114: 2017Q1.
1.137     gdt       115: 
1.104     gdt       116: xenkernel42 provides Xen 4.2.  It is no longer maintained by Xen, but
1.133     gdt       117: as of 2016-12 received backported security patches.  Xen 4.2 runs on
                    118: i386 PAE and amd64 hardware.  The only reason to run this is if you
                    119: need to use xm instead of xl, or if you need to run on hardware that
                    120: supports i386 but not amd64.  (This might also be useful if you need
                    121: an i386 dom0, if it turns out that an amd64 Xen kernel and an i386
                    122: dom0 is problematic.)
1.104     gdt       123: 
1.114     gdt       124: xenkernel45 provides Xen 4.5.  As of 2016-12, security patches were
1.133     gdt       125: released by Xen and applied to pkgsrc.  Xen 4.5 runs on amd64 hardware
                    126: only.  While slightly old, 4.5 has been tested and run by others, so
                    127: it is the conservative choice.
1.104     gdt       128: 
1.115     gdt       129: xenkernel46 provides Xen 4.6.  It is new to pkgsrc as of 2016-05.  As
1.114     gdt       130: of 2016-12, security patches were released by Xen and applied to
1.133     gdt       131: pkgsrc.  Xen 4.6 runs on amd64 hardware only For new installations,
                    132: 4.6 is probably the appropriate choice and it will likely soon be the
1.136     gdt       133: standard approach.  (If using Ubuntu guests, be sure to have the
                    134: xentools46 from December, 2016).
1.85      gdt       135: 
1.114     gdt       136: Xen 4.7 (released 2016-06) and 4.8 (released 2016-12) are not yet in
                    137: pkgsrc.
1.113     gdt       138: 
1.96      gdt       139: See also the [Xen Security Advisory page](http://xenbits.xen.org/xsa/).
                    140: 
1.85      gdt       141: Note that NetBSD support is called XEN3.  It works with Xen 3 and Xen
                    142: 4 because the hypercall interface has been stable.
1.20      gdt       143: 
1.19      gdt       144: Xen command program
                    145: -------------------
                    146: 
1.79      gdt       147: Early Xen used a program called xm to manipulate the system from the
1.19      gdt       148: dom0.  Starting in 4.1, a replacement program with similar behavior
1.79      gdt       149: called xl is provided, but it does not work well in 4.1.  In 4.2, both
1.127     gdt       150: xm and xl work fine.  4.4 is the last version that has xm.
                    151: 
                    152: You must make a global choice to use xm or xl, because it affects not
                    153: only which command you use, but the command used by rc.d scripts
                    154: (specifically xendomains) and which daemons should be run.  The
                    155: xentools packages provide xm for 3.1, 3.3 and 4.1 and xl for 4.2 and up.
                    156: 
                    157: In 4.2, you can choose to use xm by simply changing the ctl_command
1.135     gdt       158: variable and setting xend=YES in rc.conf.
1.127     gdt       159: 
                    160: With xl, virtual devices are configured in parallel, which can cause
                    161: problems if they are written assuming serial operation (e.g., updating
1.130     gdt       162: firewall rules without explicit locking).  There is now locking for
                    163: the provided scripts, which works for normal casses (e.g, file-backed
                    164: xbd, where a vnd must be allocated).  But, as of 201612, it has not
                    165: been adequately tested for a complex custom setup with a large number
                    166: of interfaces.
1.19      gdt       167: 
1.15      gdt       168: NetBSD
                    169: ------
                    170: 
1.105     gdt       171: The netbsd-6, netbsd-7, and -current branches are all reasonable
                    172: choices, with more or less the same considerations for non-Xen use.
                    173: Therefore, netbsd-7 is recommended as the stable version of the most
1.117     gdt       174: recent release for production use.  In addition, netbsd-7 and -current
                    175: have a important scheduler fix (in November of 2015) affecting
                    176: contention between dom0 and domUs; see
                    177: https://releng.netbsd.org/cgi-bin/req-7.cgi?show=1040 for a
                    178: description.  For those wanting to learn Xen or without production
                    179: stability concerns, netbsd-7 is still likely most appropriate, but
                    180: -current is also a reasonable choice.  (Xen runs ok on netbsd-5, but
                    181: the xentools packages are likely difficult to build, and netbsd-5 is
                    182: not supported.)
1.15      gdt       183: 
                    184: As of NetBSD 6, a NetBSD domU will support multiple vcpus.  There is
                    185: no SMP support for NetBSD as dom0.  (The dom0 itself doesn't really
1.105     gdt       186: need SMP for dom0 functions; the lack of support is really a problem
                    187: when using a dom0 as a normal computer.)
1.15      gdt       188: 
1.18      gdt       189: Architecture
                    190: ------------
                    191: 
1.133     gdt       192: Xen itself can run on i386 (Xen < 4.2) or amd64 hardware (all Xen
1.105     gdt       193: versions).  (Practically, almost any computer where one would want to
                    194: run Xen today supports amd64.)
1.99      gdt       195: 
1.133     gdt       196: Xen, the dom0 system, and each domU system can be either i386 or
                    197: amd64.  When building a xenkernel package, one obtains an i386 Xen
1.134     wiz       198: kernel on an i386 host, and an amd64 Xen kernel on an amd64 host.  If
1.133     gdt       199: the Xen kernel is i386, then the dom0 kernel and all domU kernels must
                    200: be i386.  With an amd64 Xen kernel, an amd64 dom0 kernel is known to
                    201: work, and an i386 dom0 kernel should in theory work.  An amd64
                    202: Xen/dom0 is known to support both i386 and amd64 domUs.
                    203: 
                    204: i386 dom0 and domU kernels must be PAE (except for an i386 Xen 3.1
1.135     gdt       205: kernel, where one can use non-PAE for dom0 and all domUs); PAE kernels
                    206: are included in the NetBSD default build.  (Note that emacs (at least)
                    207: fails if run on i386 with PAE when built without, and vice versa,
                    208: presumably due to bugs in the undump code.)
                    209: 
                    210: Because of the above, the standard approach is to use an amd64 Xen
                    211: kernel and NetBSD/amd64 for the dom0.  For domUs, NetBSD/i386 (with
                    212: the PAE kernel) and NetBSD/amd64 are in widespread use, and there is
                    213: little to no Xen-specific reason to prefer one over the other.
1.133     gdt       214: 
                    215: Note that to use an i386 dom0 with Xen 4.5 or higher, one must build
1.135     gdt       216: (or obtain from pre-built packages) an amd64 Xen kernel and install
                    217: that on the system.  (One must also use a PAE i386 kernel, but this is
                    218: also required with an i386 Xen kernel.).  Almost no one in the
                    219: NetBSD/Xen community does this, and the standard, well-tested,
                    220: approach is to use an amd64 dom0.
                    221: 
                    222: A [posting on
                    223: xen-devel](https://lists.xen.org/archives/html/xen-devel/2012-07/msg00085.html)
                    224: explained that PV system call overhead was higher on amd64, and thus
                    225: there is some notion that i386 guests are faster.  It goes on to
                    226: caution that the total situation is complex and not entirely
                    227: understood. On top of that caution, the post is about Linux, not
                    228: NetBSD.  TODO: Include link to benchmarks, if someone posts them.
1.29      gdt       229: 
1.89      gdt       230: Stability
                    231: ---------
                    232: 
                    233: Mostly, NetBSD as a dom0 or domU is quite stable.
                    234: However, there are some open PRs indicating problems.
                    235: 
1.91      gdt       236:  - [PR 48125](http://gnats.netbsd.org/48125)
                    237:  - [PR 47720](http://gnats.netbsd.org/47720)
1.89      gdt       238: 
                    239: Note also that there are issues with sparse vnd(4) instances, but
1.105     gdt       240: these are not about Xen -- they just are noticed with sparse vnd(4)
                    241: instances in support of virtual disks in a dom0.
1.89      gdt       242: 
1.15      gdt       243: Recommendation
                    244: --------------
                    245: 
1.105     gdt       246: Therefore, this HOWTO recommends running xenkernel45 or xenkernel46,
                    247: xl, the NetBSD 7 stable branch, and to use an amd64 kernel as the
                    248: dom0.  Either the i386PAE or amd64 version of NetBSD may be used as
                    249: domUs.
1.15      gdt       250: 
1.136     gdt       251: Because bugs are fixed quite often, and because of Xen security
                    252: advisories, it is good to stay up to date with NetBSD (tracking a
                    253: stable branch), with the Xen kernel (tracking a Xen version via
                    254: pkgsrc), and with the Xen tools.  Specifically, NetBSD (-7 and
                    255: -current) got an important fix affecting dom0/domU timesharing in
                    256: November, 2015, and xentools46 got a fix to enable Ubuntu guests to
                    257: boot in December, 2016.
                    258: 
1.125     gdt       259: Status
                    260: ------
1.36      gdt       261: 
1.120     gdt       262: Ideally, all versions of Xen in pkgsrc would build on all supported
1.121     gdt       263: versions of NetBSD/amd64, to the point where this section would be
                    264: silly.  However, that has not always been the case.  Besides aging
                    265: code and aging compilers, qemu (included in xentools for HVM support)
1.124     gdt       266: is difficult to build.  Note that there is intentionally no data for
1.125     gdt       267: 4.5+ up for i386, and often omits xentools info if the corresponding
                    268: kernel fails.
                    269: 
                    270: The following table gives status, with the date last checked
                    271: (generally on the most recent quarterly branch).  The first code is
                    272: "builds" if it builds ok, and "FAIL" for a failure to build.  The
                    273: second code/date only appears for xenkernel* and is "works" if it runs
1.135     gdt       274: ok as a dom0 and can support a domU, and "FAIL" if it won't boot or
                    275: run a domU.
1.124     gdt       276: 
                    277:        xenkernel3 netbsd-6 i386 FAIL 201612
                    278:        xenkernel33 netbsd-6 i386 FAIL 201612
1.125     gdt       279:        xenkernel41 netbsd-6 i386 builds 201612
                    280:        xenkernel42 netbsd-6 i386 builds 201612
1.126     gdt       281:        xentools3 netbsd-6 i386 FAIL 201612
1.128     gdt       282:        xentools33 netbsd-6 i386 FAIL 201612
1.125     gdt       283:        xentools41 netbsd-6 i386 builds 201612
1.126     gdt       284:        xentools42 netbsd-6 i386 FAIL 201612
1.64      gdt       285: 
1.124     gdt       286:        xenkernel3 netbsd-7 i386 FAIL 201412
                    287:        xenkernel33 netbsd-7 i386 FAIL 201412
1.125     gdt       288:        xenkernel41 netbsd-7 i386 builds 201412
                    289:        xenkernel42 netbsd-7 i386 builds 201412
                    290:        xentools41 netbsd-7 i386 builds 201412
1.122     gdt       291:        xentools42 netbsd-7 i386 ??FAIL 201412
1.121     gdt       292: 
1.124     gdt       293:        xenkernel3 netbsd-6 amd64 FAIL 201612
                    294:        xenkernel33 netbsd-6 amd64 FAIL 201612
1.125     gdt       295:        xenkernel41 netbsd-6 amd64 builds 201612 works 201612
1.126     gdt       296:        xenkernel42 netbsd-6 amd64 builds 201612 works 201612
1.125     gdt       297:        xenkernel45 netbsd-6 amd64 builds 201612
                    298:        xenkernel46 netbsd-6 amd64 builds 201612
                    299:        xentools41 netbsd-6 amd64 builds 201612
                    300:        xentools42 netbsd-6 amd64 builds 201612
                    301:        xentools45 netbsd-6 amd64 builds 201612
1.124     gdt       302:        xentools46 netbsd-6 amd64 FAIL 201612
1.121     gdt       303: 
1.125     gdt       304:        xenkernel3 netbsd-7 amd64 builds 201612
                    305:        xenkernel33 netbsd-7 amd64 builds 201612
                    306:        xenkernel41 netbsd-7 amd64 builds 201612
                    307:        xenkernel42 netbsd-7 amd64 builds 201612
                    308:        xenkernel45 netbsd-7 amd64 builds 201612
                    309:        xenkernel46 netbsd-7 amd64 builds 201612
                    310:        xentools3 netbsd-7 amd64 builds 201612
                    311:        xentools3-hvm netbsd-7 amd64 builds 201612
1.121     gdt       312:        xentools33 netbsd-7 amd64 FAIL 201612
1.125     gdt       313:        xentools41 netbsd-7 amd64 builds 201612
                    314:        xentools42 netbsd-7 amd64 builds 201612
                    315:        xentools45 netbsd-7 amd64 builds 201612
                    316:        xentools46 netbsd-7 amd64 builds 201612
1.69      gdt       317: 
1.15      gdt       318: NetBSD as a dom0
                    319: ================
                    320: 
                    321: NetBSD can be used as a dom0 and works very well.  The following
                    322: sections address installation, updating NetBSD, and updating Xen.
1.19      gdt       323: Note that it doesn't make sense to talk about installing a dom0 OS
                    324: without also installing Xen itself.  We first address installing
                    325: NetBSD, which is not yet a dom0, and then adding Xen, pivoting the
                    326: NetBSD install to a dom0 install by just changing the kernel and boot
                    327: configuration.
1.15      gdt       328: 
1.45      gdt       329: For experimenting with Xen, a machine with as little as 1G of RAM and
                    330: 100G of disk can work.  For running many domUs in productions, far
1.135     gdt       331: more will be needed; e.g. 4-8G and 1T of disk is reasonable for a
                    332: half-dozen domUs of 512M and 32G each.  Basically, the RAM and disk
                    333: have to be bigger than the sum of the RAM/disk needs of the dom0 and
                    334: all the domUs.
1.45      gdt       335: 
1.15      gdt       336: Styles of dom0 operation
                    337: ------------------------
                    338: 
                    339: There are two basic ways to use Xen.  The traditional method is for
                    340: the dom0 to do absolutely nothing other than providing support to some
                    341: number of domUs.  Such a system was probably installed for the sole
                    342: purpose of hosting domUs, and sits in a server room on a UPS.
                    343: 
                    344: The other way is to put Xen under a normal-usage computer, so that the
                    345: dom0 is what the computer would have been without Xen, perhaps a
                    346: desktop or laptop.  Then, one can run domUs at will.  Purists will
                    347: deride this as less secure than the previous approach, and for a
                    348: computer whose purpose is to run domUs, they are right.  But Xen and a
1.93      gdt       349: dom0 (without domUs) is not meaningfully less secure than the same
1.15      gdt       350: things running without Xen.  One can boot Xen or boot regular NetBSD
                    351: alternately with little problems, simply refraining from starting the
                    352: Xen daemons when not running Xen.
                    353: 
                    354: Note that NetBSD as dom0 does not support multiple CPUs.  This will
1.51      gdt       355: limit the performance of the Xen/dom0 workstation approach.  In theory
                    356: the only issue is that the "backend drivers" are not yet MPSAFE:
1.140   ! khorben   357:   https://mail-index.netbsd.org/netbsd-users/2014/08/29/msg015195.html
1.15      gdt       358: 
1.19      gdt       359: Installation of NetBSD
                    360: ----------------------
1.13      gdt       361: 
1.19      gdt       362: First,
1.27      jnemeth   363: [install NetBSD/amd64](/guide/inst/)
1.19      gdt       364: just as you would if you were not using Xen.
                    365: However, the partitioning approach is very important.
                    366: 
                    367: If you want to use RAIDframe for the dom0, there are no special issues
                    368: for Xen.  Typically one provides RAID storage for the dom0, and the
1.22      gdt       369: domU systems are unaware of RAID.  The 2nd-stage loader bootxx_* skips
1.111     wiz       370: over a RAID1 header to find /boot from a file system within a RAID
1.22      gdt       371: partition; this is no different when booting Xen.
1.19      gdt       372: 
                    373: There are 4 styles of providing backing storage for the virtual disks
1.93      gdt       374: used by domUs: raw partitions, LVM, file-backed vnd(4), and SAN.
1.19      gdt       375: 
                    376: With raw partitions, one has a disklabel (or gpt) partition sized for
                    377: each virtual disk to be used by the domU.  (If you are able to predict
                    378: how domU usage will evolve, please add an explanation to the HOWTO.
                    379: Seriously, needs tend to change over time.)
                    380: 
1.27      jnemeth   381: One can use [lvm(8)](/guide/lvm/) to create logical devices to use
                    382: for domU disks.  This is almost as efficient as raw disk partitions
                    383: and more flexible.  Hence raw disk partitions should typically not
                    384: be used.
1.19      gdt       385: 
1.111     wiz       386: One can use files in the dom0 file system, typically created by dd'ing
1.19      gdt       387: /dev/zero to create a specific size.  This is somewhat less efficient,
                    388: but very convenient, as one can cp the files for backup, or move them
                    389: between dom0 hosts.
                    390: 
                    391: Finally, in theory one can place the files backing the domU disks in a
                    392: SAN.  (This is an invitation for someone who has done this to add a
                    393: HOWTO page.)
1.1       mspo      394: 
1.19      gdt       395: Installation of Xen
                    396: -------------------
1.1       mspo      397: 
1.20      gdt       398: In the dom0, install sysutils/xenkernel42 and sysutils/xentools42 from
1.127     gdt       399: pkgsrc (or another matching pair).  See [the pkgsrc
1.140   ! khorben   400: documentation](https://www.NetBSD.org/docs/pkgsrc/) for help with
1.127     gdt       401: pkgsrc.  Ensure that your packages are recent; the HOWTO does not
                    402: contemplate old builds.
                    403: 
1.20      gdt       404: 
                    405: For Xen 3.1, support for HVM guests is in sysutils/xentool3-hvm.  More
                    406: recent versions have HVM support integrated in the main xentools
                    407: package.  It is entirely reasonable to run only PV guests.
                    408: 
                    409: Next you need to install the selected Xen kernel itself, which is
                    410: installed by pkgsrc as "/usr/pkg/xen*-kernel/xen.gz".  Copy it to /.
                    411: For debugging, one may copy xen-debug.gz; this is conceptually similar
                    412: to DIAGNOSTIC and DEBUG in NetBSD.  xen-debug.gz is basically only
                    413: useful with a serial console.  Then, place a NetBSD XEN3_DOM0 kernel
                    414: in /, copied from releasedir/amd64/binary/kernel/netbsd-XEN3_DOM0.gz
1.75      gdt       415: of a NetBSD build.  If using i386, use
                    416: releasedir/i386/binary/kernel/netbsd-XEN3PAE_DOM0.gz.  (If using Xen
                    417: 3.1 and i386, you may use XEN3_DOM0 with the non-PAE Xen.  But you
                    418: should not use Xen 3.1.)  Both xen and the NetBSD kernel may be (and
                    419: typically are) left compressed.
                    420: 
1.135     gdt       421: In a dom0, kernfs is mandatory for xend to communicate with the
                    422: kernel, so ensure that /kern is in fstab.  (A standard NetBSD install
                    423: should already mount /kern.)
1.20      gdt       424: 
                    425: Because you already installed NetBSD, you have a working boot setup
                    426: with an MBR bootblock, either bootxx_ffsv1 or bootxx_ffsv2 at the
1.135     gdt       427: beginning of your root file system, have /boot, and likely also
1.20      gdt       428: /boot.cfg.  (If not, fix before continuing!)
                    429: 
1.76      gdt       430: Add a line to to /boot.cfg to boot Xen.  See boot.cfg(5) for an
                    431: example.  The basic line is
1.20      gdt       432: 
1.37      gdt       433:         menu=Xen:load /netbsd-XEN3_DOM0.gz console=pc;multiboot /xen.gz dom0_mem=256M
1.20      gdt       434: 
                    435: which specifies that the dom0 should have 256M, leaving the rest to be
1.77      gdt       436: allocated for domUs.  To use a serial console, use
                    437: 
1.139     gson      438:         menu=Xen:load /netbsd-XEN3_DOM0.gz;multiboot /xen.gz dom0_mem=256M console=com1 com1=9600,8n1
1.77      gdt       439: 
                    440: which will use the first serial port for Xen (which counts starting
1.139     gson      441: from 1, unlike NetBSD which counts starting from 0), forcing
                    442: speed/parity.  Because the NetBSD command line lacks a
                    443: "console=pc" argument, it will use the default "xencons" console device,
                    444: which directs the console I/O through Xen to the same console device Xen
                    445: itself uses (in this case, the serial port).
                    446: 
                    447: In an attempt to add performance, one can also add
1.37      gdt       448: 
                    449:         dom0_max_vcpus=1 dom0_vcpus_pin
                    450: 
                    451: to force only one vcpu to be provided (since NetBSD dom0 can't use
1.111     wiz       452: more) and to pin that vcpu to a physical CPU.  TODO: benchmark this.
1.20      gdt       453: 
1.93      gdt       454: Xen has [many boot
                    455: options](http://xenbits.xenproject.org/docs/4.5-testing/misc/xen-command-line.html),
1.111     wiz       456: and other than dom0 memory and max_vcpus, they are generally not
1.93      gdt       457: necessary.
                    458: 
1.20      gdt       459: As with non-Xen systems, you should have a line to boot /netbsd (a
1.127     gdt       460: kernel that works without Xen).  Consider a line to boot /netbsd.ok (a
                    461: fallback version of the non-Xen kernel, updated manually when you are
                    462: sure /netbsd is ok).  Consider also a line to boot fallback versions
                    463: of Xen and the dom0 kernel, but note that non-Xen NetBSD can be used
                    464: to resolve Xen booting issues.
                    465: 
                    466: Probably you want a default=N line to choose Xen in the absence of
                    467: intervention.
1.1       mspo      468: 
1.76      gdt       469: Now, reboot so that you are running a DOM0 kernel under Xen, rather
                    470: than GENERIC without Xen.
                    471: 
1.54      gdt       472: Using grub (historic)
                    473: ---------------------
                    474: 
                    475: Before NetBSD's native bootloader could support Xen, the use of
                    476: grub was recommended.  If necessary, see the
1.135     gdt       477: [old grub information](/ports/xen/howto-grub).
1.54      gdt       478: 
1.28      gdt       479: The [HowTo on Installing into
1.140   ! khorben   480: RAID-1](https://mail-index.NetBSD.org/port-xen/2006/03/01/0010.html)
1.28      gdt       481: explains how to set up booting a dom0 with Xen using grub with
                    482: NetBSD's RAIDframe.  (This is obsolete with the use of NetBSD's native
1.135     gdt       483: boot.  Now, just create a system with RAID-1, and alter /boot.cfg as
                    484: described above.)
1.28      gdt       485: 
1.21      gdt       486: Configuring Xen
                    487: ---------------
                    488: 
1.53      gdt       489: Xen logs will be in /var/log/xen.
                    490: 
1.76      gdt       491: Now, you have a system that will boot Xen and the dom0 kernel, but not
                    492: do anything else special.  Make sure that you have rebooted into Xen.
                    493: There will be no domUs, and none can be started because you still have
1.102     gdt       494: to configure the dom0 daemons.
1.21      gdt       495: 
1.102     gdt       496: The daemons which should be run vary with Xen version and with whether
1.127     gdt       497: one is using xm or xl.  The Xen 3.1, 3.3 and 4.1 packages use xm.  Xen
                    498: 4.2 and up packages use xl.  To use xm with 4.2, edit xendomains to
                    499: use xm instead.
                    500: 
1.132     gdt       501: For 3.1 and 3.3, you should enable xend and xenbackendd:
1.31      gdt       502: 
1.32      gdt       503:         xend=YES
                    504:         xenbackendd=YES
1.31      gdt       505: 
1.132     gdt       506: For 4.1 and up, you should enable xencommons.  Not enabling xencommons
                    507: will result in a hang; it is necessary to hit ^C on the console to let
                    508: the machine finish booting.  If you are using xm (default in 4.1, or
                    509: if you changed xendomains in 4.2), you should also enable xend:
1.31      gdt       510: 
1.132     gdt       511:         xend=YES # only if using xm, and only installed <= 4.2
1.53      gdt       512:         xencommons=YES
1.31      gdt       513: 
                    514: TODO: Recommend for/against xen-watchdog.
1.27      jnemeth   515: 
1.53      gdt       516: After you have configured the daemons and either started them (in the
1.79      gdt       517: order given) or rebooted, use xm or xl to inspect Xen's boot messages,
1.102     gdt       518: available resources, and running domains.  An example with xl follows:
1.34      gdt       519: 
1.102     gdt       520:         # xl dmesg
1.43      gdt       521:        [xen's boot info]
1.102     gdt       522:         # xl info
1.43      gdt       523:        [available memory, etc.]
1.102     gdt       524:         # xl list
1.43      gdt       525:         Name              Id  Mem(MB)  CPU  State  Time(s)  Console
                    526:         Domain-0           0       64    0  r----     58.1
1.33      gdt       527: 
1.88      gdt       528: ### Issues with xencommons
                    529: 
                    530: xencommons starts xenstored, which stores data on behalf of dom0 and
                    531: domUs.  It does not currently work to stop and start xenstored.
                    532: Certainly all domUs should be shutdown first, following the sort order
                    533: of the rc.d scripts.  However, the dom0 sets up state with xenstored,
                    534: and is not notified when xenstored exits, leading to not recreating
                    535: the state when the new xenstored starts.  Until there's a mechanism to
                    536: make this work, one should not expect to be able to restart xenstored
                    537: (and thus xencommons).  There is currently no reason to expect that
                    538: this will get fixed any time soon.
                    539: 
1.127     gdt       540: ### No-longer needed advice about devices
                    541: 
                    542: The installation of NetBSD should already have created devices for xen
                    543: (xencons, xenevt, xsd_kva), but if they are not present, create them:
                    544: 
                    545:         cd /dev && sh MAKEDEV xen
                    546: 
1.41      gdt       547: anita (for testing NetBSD)
                    548: --------------------------
                    549: 
1.82      gdt       550: With the setup so far (assuming 4.2/xl), one should be able to run
                    551: anita (see pkgsrc/misc/py-anita) to test NetBSD releases, by doing (as
                    552: root, because anita must create a domU):
                    553: 
                    554:         anita --vmm=xl test file:///usr/obj/i386/
                    555: 
                    556: Alternatively, one can use --vmm=xm to use xm-based domU creation
                    557: instead (and must, on Xen <= 4.1).   TODO: confirm that anita xl really works.
                    558:     
1.40      gdt       559: Xen-specific NetBSD issues
                    560: --------------------------
                    561: 
                    562: There are (at least) two additional things different about NetBSD as a
                    563: dom0 kernel compared to hardware.
                    564: 
1.111     wiz       565: One is that the module ABI is different because some of the #defines
1.109     gdt       566: change, so one must build modules for Xen.  As of netbsd-7, the build
                    567: system does this automatically.  TODO: check this.  (Before building
                    568: Xen modules was added, it was awkward to use modules to the point
                    569: where it was considered that it did not work.)
1.40      gdt       570: 
                    571: The other difference is that XEN3_DOM0 does not have exactly the same
                    572: options as GENERIC.  While it is debatable whether or not this is a
                    573: bug, users should be aware of this and can simply add missing config
                    574: items if desired.
                    575: 
1.15      gdt       576: Updating NetBSD in a dom0
                    577: -------------------------
                    578: 
                    579: This is just like updating NetBSD on bare hardware, assuming the new
                    580: version supports the version of Xen you are running.  Generally, one
                    581: replaces the kernel and reboots, and then overlays userland binaries
                    582: and adjusts /etc.
                    583: 
                    584: Note that one must update both the non-Xen kernel typically used for
                    585: rescue purposes and the DOM0 kernel used with Xen.
                    586: 
1.55      gdt       587: Converting from grub to /boot
                    588: -----------------------------
                    589: 
                    590: These instructions were [TODO: will be] used to convert a system from
                    591: grub to /boot.  The system was originally installed in February of
                    592: 2006 with a RAID1 setup and grub to boot Xen 2, and has been updated
                    593: over time.  Before these commands, it was running NetBSD 6 i386, Xen
                    594: 4.1 and grub, much like the message linked earlier in the grub
                    595: section.
                    596: 
1.111     wiz       597:         # Install MBR bootblocks on both disks. 
1.55      gdt       598:         fdisk -i /dev/rwd0d
                    599:         fdisk -i /dev/rwd1d
                    600:         # Install NetBSD primary boot loader (/ is FFSv1) into RAID1 components.
                    601:         installboot -v /dev/rwd0d /usr/mdec/bootxx_ffsv1
                    602:         installboot -v /dev/rwd1d /usr/mdec/bootxx_ffsv1
                    603:         # Install secondary boot loader
                    604:         cp -p /usr/mdec/boot /
1.111     wiz       605:         # Create boot.cfg following earlier guidance:
1.55      gdt       606:         menu=Xen:load /netbsd-XEN3PAE_DOM0.gz console=pc;multiboot /xen.gz dom0_mem=256M
                    607:         menu=Xen.ok:load /netbsd-XEN3PAE_DOM0.ok.gz console=pc;multiboot /xen.ok.gz dom0_mem=256M
                    608:         menu=GENERIC:boot
                    609:         menu=GENERIC single-user:boot -s
                    610:         menu=GENERIC.ok:boot netbsd.ok
                    611:         menu=GENERIC.ok single-user:boot netbsd.ok -s
                    612:         menu=Drop to boot prompt:prompt
                    613:         default=1
                    614:         timeout=30
                    615: 
                    616: TODO: actually do this and fix it if necessary.
1.22      gdt       617: 
1.102     gdt       618: Upgrading Xen versions
1.15      gdt       619: ---------------------
                    620: 
1.110     gdt       621: Minor version upgrades are trivial.  Just rebuild/replace the
                    622: xenkernel version and copy the new xen.gz to / (where /boot.cfg
                    623: references it), and reboot.
                    624: 
                    625: Major version upgrades are conceptually not difficult, but can run
                    626: into all the issues found when installing Xen.  Assuming migration
                    627: from 4.1 to 4.2, remove the xenkernel41 and xentools41 packages and
                    628: install the xenkernel42 and xentools42 packages.  Copy the 4.2 xen.gz
                    629: to /.
1.21      gdt       630: 
1.102     gdt       631: Ensure that the contents of /etc/rc.d/xen* are correct.  Specifically,
                    632: they must match the package you just installed and not be left over
                    633: from some previous installation.
                    634: 
                    635: Enable the correct set of daemons; see the configuring section above.
                    636: (Upgrading from 3.x to 4.x without doing this will result in a hang.)
                    637: 
                    638: Ensure that the domU config files are valid for the new version.
1.110     gdt       639: Specifically, for 4.x remove autorestart=True, and ensure that disks
                    640: are specified with numbers as the second argument, as the examples
                    641: above show, and not NetBSD device names.
1.15      gdt       642: 
1.97      gdt       643: Hardware known to work
                    644: ----------------------
                    645: 
                    646: Arguably, this section is misplaced, and there should be a page of
                    647: hardware that runs NetBSD/amd64 well, with the mostly-well-founded
                    648: assumption that NetBSD/xen runs fine on any modern hardware that
1.129     gdt       649: NetBSD/amd64 runs well on.  Until then, we give motherboard/CPU (and
                    650: sometimes RAM) pairs/triples to aid those choosing a motherboard.
                    651: Note that Xen systems usually do not run X, so a listing here does not
                    652: imply that X works at all.
1.97      gdt       653: 
                    654:         Supermicro X9SRL-F, Xeon E5-1650 v2, 96 GiB ECC
                    655:         Supermicro ??, Atom C2758 (8 core), 32 GiB ECC
                    656:         ASUS M5A78L-M/USB3 AM3+ microATX, AMD Piledriver X8 4000MHz, 16 GiB ECC
                    657: 
                    658: Older hardware:
                    659: 
1.98      gdt       660:         Intel D915GEV, Pentium4 CPU 3.40GHz, 4GB 533MHz Synchronous DDR2
1.129     gdt       661:         INTEL DG33FB, "Intel(R) Core(TM)2 Duo CPU     E6850  @ 3.00GHz"
                    662:         INTEL DG33FB, "Intel(R) Core(TM)2 Duo CPU     E8400  @ 3.00GHz"
1.28      gdt       663: 
1.82      gdt       664: Running Xen under qemu
                    665: ----------------------
                    666: 
                    667: The astute reader will note that this section is somewhat twisted.
                    668: However, it can be useful to run Xen under qemu either because the
                    669: version of NetBSD as a dom0 does not run on the hardware in use, or to
                    670: generate automated test cases involving Xen.
                    671: 
1.84      gdt       672: In 2015-01, the following combination was reported to mostly work:
1.82      gdt       673: 
                    674:         host OS: NetBSD/amd64 6.1.4
                    675:         qemu: 2.2.0 from pkgsrc
                    676:         Xen kernel: xenkernel42-4.2.5nb1 from pkgsrc
                    677:         dom0 kernel: NetBSD/amd64 6.1.5
                    678:         Xen tools: xentools42-4.2.5 from pkgsrc
                    679: 
1.140   ! khorben   680: See [PR 47720](https://gnats.netbsd.org/47720) for a problem with dom0
1.91      gdt       681: shutdown.
1.84      gdt       682: 
1.28      gdt       683: Unprivileged domains (domU)
                    684: ===========================
                    685: 
                    686: This section describes general concepts about domUs.  It does not
1.33      gdt       687: address specific domU operating systems or how to install them.  The
                    688: config files for domUs are typically in /usr/pkg/etc/xen, and are
1.60      wiki      689: typically named so that the file name, domU name and the domU's host
1.33      gdt       690: name match.
                    691: 
1.111     wiz       692: The domU is provided with CPU and memory by Xen, configured by the
1.33      gdt       693: dom0.  The domU is provided with disk and network by the dom0,
                    694: mediated by Xen, and configured in the dom0.
                    695: 
                    696: Entropy in domUs can be an issue; physical disks and network are on
                    697: the dom0.  NetBSD's /dev/random system works, but is often challenged.
                    698: 
1.48      gdt       699: Config files
                    700: ------------
                    701: 
                    702: There is no good order to present config files and the concepts
                    703: surrounding what is being configured.  We first show an example config
                    704: file, and then in the various sections give details.
                    705: 
                    706: See (at least in xentools41) /usr/pkg/share/examples/xen/xmexample*,
                    707: for a large number of well-commented examples, mostly for running
                    708: GNU/Linux.
                    709: 
                    710: The following is an example minimal domain configuration file
                    711: "/usr/pkg/etc/xen/foo".  It is (with only a name change) an actual
                    712: known working config file on Xen 4.1 (NetBSD 5 amd64 dom0 and NetBSD 5
                    713: i386 domU).  The domU serves as a network file server.
                    714: 
                    715:         # -*- mode: python; -*-
                    716: 
                    717:         kernel = "/netbsd-XEN3PAE_DOMU-i386-foo.gz"
                    718:         memory = 1024
                    719:         vif = [ 'mac=aa:00:00:d1:00:09,bridge=bridge0' ]
                    720:         disk = [ 'file:/n0/xen/foo-wd0,0x0,w',
                    721:                  'file:/n0/xen/foo-wd1,0x1,w' ]
                    722: 
                    723: The domain will have the same name as the file.  The kernel has the
                    724: host/domU name in it, so that on the dom0 one can update the various
                    725: domUs independently.  The vif line causes an interface to be provided,
                    726: with a specific mac address (do not reuse MAC addresses!), in bridge
                    727: mode.  Two disks are provided, and they are both writable; the bits
                    728: are stored in files and Xen attaches them to a vnd(4) device in the
1.111     wiz       729: dom0 on domain creation.  The system treats xbd0 as the boot device
1.48      gdt       730: without needing explicit configuration.
                    731: 
                    732: By default xm looks for domain config files in /usr/pkg/etc/xen.  Note
                    733: that "xm create" takes the name of a config file, while other commands
                    734: take the name of a domain.  To create the domain, connect to the
                    735: console, create the domain while attaching the console, shutdown the
                    736: domain, and see if it has finished stopping, do (or xl with Xen >=
                    737: 4.2):
                    738: 
                    739:         xm create foo
                    740:         xm console foo
                    741:         xm create -c foo
                    742:         xm shutdown foo
1.90      gdt       743:         xm list
1.48      gdt       744: 
                    745: Typing ^] will exit the console session.  Shutting down a domain is
                    746: equivalent to pushing the power button; a NetBSD domU will receive a
                    747: power-press event and do a clean shutdown.  Shutting down the dom0
                    748: will trigger controlled shutdowns of all configured domUs.
                    749: 
                    750: domU kernels
                    751: ------------
                    752: 
                    753: On a physical computer, the BIOS reads sector 0, and a chain of boot
                    754: loaders finds and loads a kernel.  Normally this comes from the root
1.111     wiz       755: file system.  With Xen domUs, the process is totally different.  The
1.48      gdt       756: normal path is for the domU kernel to be a file in the dom0's
1.111     wiz       757: file system.  At the request of the dom0, Xen loads that kernel into a
1.48      gdt       758: new domU instance and starts execution.  While domU kernels can be
                    759: anyplace, reasonable places to store domU kernels on the dom0 are in /
                    760: (so they are near the dom0 kernel), in /usr/pkg/etc/xen (near the
                    761: config files), or in /u0/xen (where the vdisks are).
                    762: 
1.59      gdt       763: Note that loading the domU kernel from the dom0 implies that boot
                    764: blocks, /boot, /boot.cfg, and so on are all ignored in the domU.
1.48      gdt       765: See the VPS section near the end for discussion of alternate ways to
                    766: obtain domU kernels.
                    767: 
1.33      gdt       768: CPU and memory
                    769: --------------
                    770: 
1.48      gdt       771: A domain is provided with some number of vcpus, less than the number
1.111     wiz       772: of CPUs seen by the hypervisor.  (For a dom0, this is controlled by
1.48      gdt       773: the boot argument "dom0_max_vcpus=1".)  For a domU, it is controlled
                    774: from the config file by the "vcpus = N" directive.
                    775: 
                    776: A domain is provided with memory; this is controlled in the config
                    777: file by "memory = N" (in megabytes).  In the straightforward case, the
                    778: sum of the the memory allocated to the dom0 and all domUs must be less
1.33      gdt       779: than the available memory.
                    780: 
                    781: Xen also provides a "balloon" driver, which can be used to let domains
                    782: use more memory temporarily.  TODO: Explain better, and explain how
                    783: well it works with NetBSD.
1.28      gdt       784: 
                    785: Virtual disks
                    786: -------------
                    787: 
1.33      gdt       788: With the file/vnd style, typically one creates a directory,
                    789: e.g. /u0/xen, on a disk large enough to hold virtual disks for all
                    790: domUs.  Then, for each domU disk, one writes zeros to a file that then
                    791: serves to hold the virtual disk's bits; a suggested name is foo-xbd0
                    792: for the first virtual disk for the domU called foo.  Writing zeros to
                    793: the file serves two purposes.  One is that preallocating the contents
                    794: improves performance.  The other is that vnd on sparse files has
                    795: failed to work.  TODO: give working/notworking NetBSD versions for
1.127     gdt       796: sparse vnd and gnats reference.  Note that the use of file/vnd for Xen
                    797: is not really different than creating a file-backed virtual disk for
                    798: some other purpose, except that xentools handles the vnconfig
                    799: commands.  To create an empty 4G virtual disk, simply do
1.39      gdt       800: 
                    801:         dd if=/dev/zero of=foo-xbd0 bs=1m count=4096
1.33      gdt       802: 
1.89      gdt       803: Do not use qemu-img-xen, because this will create sparse file.  There
                    804: have been recent (2015) reports of sparse vnd(4) devices causing
                    805: lockups, but there is apparently no PR.
                    806: 
1.33      gdt       807: With the lvm style, one creates logical devices.  They are then used
1.48      gdt       808: similarly to vnds.  TODO: Add an example with lvm.
                    809: 
                    810: In domU config files, the disks are defined as a sequence of 3-tuples.
                    811: The first element is "method:/path/to/disk".  Common methods are
                    812: "file:" for file-backed vnd. and "phy:" for something that is already
                    813: a (TODO: character or block) device.
                    814: 
                    815: The second element is an artifact of how virtual disks are passed to
                    816: Linux, and a source of confusion with NetBSD Xen usage.  Linux domUs
                    817: are given a device name to associate with the disk, and values like
                    818: "hda1" or "sda1" are common.  In a NetBSD domU, the first disk appears
                    819: as xbd0, the second as xbd1, and so on.  However, xm/xl demand a
                    820: second argument.  The name given is converted to a major/minor by
1.49      gdt       821: calling stat(2) on the name in /dev and this is passed to the domU.
                    822: In the general case, the dom0 and domU can be different operating
1.48      gdt       823: systems, and it is an unwarranted assumption that they have consistent
                    824: numbering in /dev, or even that the dom0 OS has a /dev.  With NetBSD
                    825: as both dom0 and domU, using values of 0x0 for the first disk and 0x1
1.49      gdt       826: for the second works fine and avoids this issue.  For a GNU/Linux
                    827: guest, one can create /dev/hda1 in /dev, or to pass 0x301 for
                    828: /dev/hda1.
1.48      gdt       829: 
                    830: The third element is "w" for writable disks, and "r" for read-only
                    831: disks.
1.28      gdt       832: 
1.127     gdt       833: Note that NetBSD by default creates only vnd[0123].  If you need more
                    834: than 4 total virtual disks at a time, run e.g. "./MAKEDEV vnd4" in the
                    835: dom0.
                    836: 
                    837: Note that NetBSD by default creates only xbd[0123].  If you need more
                    838: virtual disks in a domU, run e.g. "./MAKEDEV xbd4" in the domU.
                    839: 
1.28      gdt       840: Virtual Networking
                    841: ------------------
                    842: 
1.111     wiz       843: Xen provides virtual Ethernets, each of which connects the dom0 and a
1.46      gdt       844: domU.  For each virtual network, there is an interface "xvifN.M" in
                    845: the dom0, and in domU index N, a matching interface xennetM (NetBSD
                    846: name).  The interfaces behave as if there is an Ethernet with two
1.111     wiz       847: adapters connected.  From this primitive, one can construct various
1.46      gdt       848: configurations.  We focus on two common and useful cases for which
                    849: there are existing scripts: bridging and NAT.
1.28      gdt       850: 
1.48      gdt       851: With bridging (in the example above), the domU perceives itself to be
                    852: on the same network as the dom0.  For server virtualization, this is
                    853: usually best.  Bridging is accomplished by creating a bridge(4) device
                    854: and adding the dom0's physical interface and the various xvifN.0
                    855: interfaces to the bridge.  One specifies "bridge=bridge0" in the domU
                    856: config file.  The bridge must be set up already in the dom0; an
                    857: example /etc/ifconfig.bridge0 is:
1.46      gdt       858: 
                    859:         create
                    860:         up
                    861:         !brconfig bridge0 add wm0
1.28      gdt       862: 
                    863: With NAT, the domU perceives itself to be behind a NAT running on the
                    864: dom0.  This is often appropriate when running Xen on a workstation.
1.48      gdt       865: TODO: NAT appears to be configured by "vif = [ '' ]".
1.28      gdt       866: 
1.49      gdt       867: The MAC address specified is the one used for the interface in the new
1.53      gdt       868: domain.  The interface in dom0 will use this address XOR'd with
1.49      gdt       869: 00:00:00:01:00:00.  Random MAC addresses are assigned if not given.
                    870: 
1.33      gdt       871: Sizing domains
                    872: --------------
                    873: 
                    874: Modern x86 hardware has vast amounts of resources.  However, many
                    875: virtual servers can function just fine on far less.  A system with
                    876: 256M of RAM and a 4G disk can be a reasonable choice.  Note that it is
                    877: far easier to adjust virtual resources than physical ones.  For
                    878: memory, it's just a config file edit and a reboot.  For disk, one can
                    879: create a new file and vnconfig it (or lvm), and then dump/restore,
                    880: just like updating physical disks, but without having to be there and
                    881: without those pesky connectors.
                    882: 
1.48      gdt       883: Starting domains automatically
                    884: ------------------------------
1.28      gdt       885: 
1.48      gdt       886: To start domains foo at bar at boot and shut them down cleanly on dom0
                    887: shutdown, in rc.conf add:
1.28      gdt       888: 
1.48      gdt       889:         xendomains="foo bar"
1.28      gdt       890: 
1.111     wiz       891: Note that earlier versions of the xentools41 xendomains rc.d script
                    892: used xl, when one should use xm with 4.1.
1.28      gdt       893: 
                    894: Creating specific unprivileged domains (domU)
                    895: =============================================
1.14      gdt       896: 
                    897: Creating domUs is almost entirely independent of operating system.  We
1.49      gdt       898: have already presented the basics of config files.  Note that you must
                    899: have already completed the dom0 setup so that "xl list" (or "xm list")
                    900: works.
1.14      gdt       901: 
                    902: Creating an unprivileged NetBSD domain (domU)
                    903: ---------------------------------------------
1.1       mspo      904: 
1.49      gdt       905: See the earlier config file, and adjust memory.  Decide on how much
                    906: storage you will provide, and prepare it (file or lvm).
                    907: 
1.111     wiz       908: While the kernel will be obtained from the dom0 file system, the same
1.49      gdt       909: file should be present in the domU as /netbsd so that tools like
                    910: savecore(8) can work.   (This is helpful but not necessary.)
                    911: 
                    912: The kernel must be specifically for Xen and for use as a domU.  The
                    913: i386 and amd64 provide the following kernels:
                    914: 
                    915:         i386 XEN3_DOMU
                    916:         i386 XEN3PAE_DOMU
1.95      gdt       917:         amd64 XEN3_DOMU
1.5       mspo      918: 
1.49      gdt       919: Unless using Xen 3.1 (and you shouldn't) with i386-mode Xen, you must
                    920: use the PAE version of the i386 kernel.
                    921: 
                    922: This will boot NetBSD, but this is not that useful if the disk is
                    923: empty.  One approach is to unpack sets onto the disk outside of xen
                    924: (by mounting it, just as you would prepare a physical disk for a
                    925: system you can't run the installer on).
                    926: 
                    927: A second approach is to run an INSTALL kernel, which has a miniroot
                    928: and can load sets from the network.  To do this, copy the INSTALL
                    929: kernel to / and change the kernel line in the config file to:
1.5       mspo      930: 
1.49      gdt       931:         kernel = "/home/bouyer/netbsd-INSTALL_XEN3_DOMU"
1.5       mspo      932: 
1.49      gdt       933: Then, start the domain as "xl create -c configname".
1.1       mspo      934: 
1.49      gdt       935: Alternatively, if you want to install NetBSD/Xen with a CDROM image, the following
                    936: line should be used in the config file.
1.1       mspo      937: 
1.3       mspo      938:     disk = [ 'phy:/dev/wd0e,0x1,w', 'phy:/dev/cd0a,0x2,r' ]
1.1       mspo      939: 
                    940: After booting the domain, the option to install via CDROM may be
1.49      gdt       941: selected.  The CDROM device should be changed to `xbd1d`.
1.1       mspo      942: 
1.49      gdt       943: Once done installing, "halt -p" the new domain (don't reboot or halt,
                    944: it would reload the INSTALL_XEN3_DOMU kernel even if you changed the
                    945: config file), switch the config file back to the XEN3_DOMU kernel,
                    946: and start the new domain again. Now it should be able to use "root on
                    947: xbd0a" and you should have a, functional NetBSD domU.
1.1       mspo      948: 
1.49      gdt       949: TODO: check if this is still accurate.
1.1       mspo      950: When the new domain is booting you'll see some warnings about *wscons*
                    951: and the pseudo-terminals. These can be fixed by editing the files
1.5       mspo      952: `/etc/ttys` and `/etc/wscons.conf`. You must disable all terminals in
                    953: `/etc/ttys`, except *console*, like this:
1.1       mspo      954: 
1.3       mspo      955:     console "/usr/libexec/getty Pc"         vt100   on secure
                    956:     ttyE0   "/usr/libexec/getty Pc"         vt220   off secure
                    957:     ttyE1   "/usr/libexec/getty Pc"         vt220   off secure
                    958:     ttyE2   "/usr/libexec/getty Pc"         vt220   off secure
                    959:     ttyE3   "/usr/libexec/getty Pc"         vt220   off secure
1.1       mspo      960: 
1.5       mspo      961: Finally, all screens must be commented out from `/etc/wscons.conf`.
1.1       mspo      962: 
                    963: It is also desirable to add
                    964: 
1.49      gdt       965:         powerd=YES
1.1       mspo      966: 
1.5       mspo      967: in rc.conf. This way, the domain will be properly shut down if
1.53      gdt       968: `xm shutdown -R` or `xm shutdown -H` is used on the dom0.
1.1       mspo      969: 
1.92      gdt       970: It is not strictly necessary to have a kernel (as /netbsd) in the domU
1.111     wiz       971: file system.  However, various programs (e.g. netstat) will use that
1.92      gdt       972: kernel to look up symbols to read from kernel virtual memory.  If
                    973: /netbsd is not the running kernel, those lookups will fail.  (This is
                    974: not really a Xen-specific issue, but because the domU kernel is
                    975: obtained from the dom0, it is far more likely to be out of sync or
                    976: missing with Xen.)
                    977: 
1.14      gdt       978: Creating an unprivileged Linux domain (domU)
1.5       mspo      979: --------------------------------------------
1.1       mspo      980: 
                    981: Creating unprivileged Linux domains isn't much different from
                    982: unprivileged NetBSD domains, but there are some details to know.
                    983: 
                    984: First, the second parameter passed to the disk declaration (the '0x1' in
                    985: the example below)
                    986: 
1.3       mspo      987:     disk = [ 'phy:/dev/wd0e,0x1,w' ]
1.1       mspo      988: 
                    989: does matter to Linux. It wants a Linux device number here (e.g. 0x300
1.49      gdt       990: for hda).  Linux builds device numbers as: (major \<\< 8 + minor).
                    991: So, hda1 which has major 3 and minor 1 on a Linux system will have
                    992: device number 0x301.  Alternatively, devices names can be used (hda,
                    993: hdb, ...)  as xentools has a table to map these names to devices
                    994: numbers.  To export a partition to a Linux guest we can use:
1.1       mspo      995: 
1.49      gdt       996:         disk = [ 'phy:/dev/wd0e,0x300,w' ]
                    997:         root = "/dev/hda1 ro"
1.1       mspo      998: 
                    999: and it will appear as /dev/hda on the Linux system, and be used as root
                   1000: partition.
                   1001: 
1.49      gdt      1002: To install the Linux system on the partition to be exported to the
                   1003: guest domain, the following method can be used: install
                   1004: sysutils/e2fsprogs from pkgsrc.  Use mke2fs to format the partition
                   1005: that will be the root partition of your Linux domain, and mount it.
                   1006: Then copy the files from a working Linux system, make adjustments in
                   1007: `/etc` (fstab, network config).  It should also be possible to extract
                   1008: binary packages such as .rpm or .deb directly to the mounted partition
                   1009: using the appropriate tool, possibly running under NetBSD's Linux
1.111     wiz      1010: emulation.  Once the file system has been populated, umount it.  If
                   1011: desirable, the file system can be converted to ext3 using tune2fs -j.
1.49      gdt      1012: It should now be possible to boot the Linux guest domain, using one of
                   1013: the vmlinuz-\*-xenU kernels available in the Xen binary distribution.
1.1       mspo     1014: 
1.111     wiz      1015: To get the Linux console right, you need to add:
1.1       mspo     1016: 
1.3       mspo     1017:     extra = "xencons=tty1"
1.1       mspo     1018: 
1.111     wiz      1019: to your configuration since not all Linux distributions auto-attach a
1.1       mspo     1020: tty to the xen console.
                   1021: 
1.14      gdt      1022: Creating an unprivileged Solaris domain (domU)
1.5       mspo     1023: ----------------------------------------------
1.1       mspo     1024: 
1.50      gdt      1025: See possibly outdated
                   1026: [Solaris domU instructions](/ports/xen/howto-solaris/).
1.5       mspo     1027: 
1.1       mspo     1028: 
1.52      gdt      1029: PCI passthrough: Using PCI devices in guest domains
                   1030: ---------------------------------------------------
1.1       mspo     1031: 
1.53      gdt      1032: The dom0 can give other domains access to selected PCI
1.52      gdt      1033: devices. This can allow, for example, a non-privileged domain to have
                   1034: access to a physical network interface or disk controller.  However,
                   1035: keep in mind that giving a domain access to a PCI device most likely
                   1036: will give the domain read/write access to the whole physical memory,
                   1037: as PCs don't have an IOMMU to restrict memory access to DMA-capable
1.53      gdt      1038: device.  Also, it's not possible to export ISA devices to non-dom0
1.52      gdt      1039: domains, which means that the primary VGA adapter can't be exported.
                   1040: A guest domain trying to access the VGA registers will panic.
                   1041: 
1.53      gdt      1042: If the dom0 is NetBSD, it has to be running Xen 3.1, as support has
1.52      gdt      1043: not been ported to later versions at this time.
                   1044: 
                   1045: For a PCI device to be exported to a domU, is has to be attached to
                   1046: the "pciback" driver in dom0.  Devices passed to the dom0 via the
                   1047: pciback.hide boot parameter will attach to "pciback" instead of the
                   1048: usual driver.  The list of devices is specified as "(bus:dev.func)",
1.5       mspo     1049: where bus and dev are 2-digit hexadecimal numbers, and func a
                   1050: single-digit number:
1.1       mspo     1051: 
1.52      gdt      1052:         pciback.hide=(00:0a.0)(00:06.0)
1.1       mspo     1053: 
1.52      gdt      1054: pciback devices should show up in the dom0's boot messages, and the
1.5       mspo     1055: devices should be listed in the `/kern/xen/pci` directory.
1.1       mspo     1056: 
1.52      gdt      1057: PCI devices to be exported to a domU are listed in the "pci" array of
                   1058: the domU's config file, with the format "0000:bus:dev.func".
1.1       mspo     1059: 
1.52      gdt      1060:         pci = [ '0000:00:06.0', '0000:00:0a.0' ]
1.1       mspo     1061: 
1.52      gdt      1062: In the domU an "xpci" device will show up, to which one or more pci
1.111     wiz      1063: buses will attach.  Then the PCI drivers will attach to PCI buses as
1.52      gdt      1064: usual.  Note that the default NetBSD DOMU kernels do not have "xpci"
                   1065: or any PCI drivers built in by default; you have to build your own
                   1066: kernel to use PCI devices in a domU.  Here's a kernel config example;
                   1067: note that only the "xpci" lines are unusual.
                   1068: 
                   1069:         include         "arch/i386/conf/XEN3_DOMU"
                   1070: 
1.111     wiz      1071:         # Add support for PCI buses to the XEN3_DOMU kernel
1.52      gdt      1072:         xpci* at xenbus ?
                   1073:         pci* at xpci ?
                   1074: 
                   1075:         # PCI USB controllers
                   1076:         uhci*   at pci? dev ? function ?        # Universal Host Controller (Intel)
                   1077: 
                   1078:         # USB bus support
                   1079:         usb*    at uhci?
                   1080: 
                   1081:         # USB Hubs
                   1082:         uhub*   at usb?
                   1083:         uhub*   at uhub? port ? configuration ? interface ?
                   1084: 
                   1085:         # USB Mass Storage
                   1086:         umass*  at uhub? port ? configuration ? interface ?
                   1087:         wd*     at umass?
                   1088:         # SCSI controllers
                   1089:         ahc*    at pci? dev ? function ?        # Adaptec [23]94x, aic78x0 SCSI
                   1090: 
                   1091:         # SCSI bus support (for both ahc and umass)
                   1092:         scsibus* at scsi?
                   1093: 
                   1094:         # SCSI devices
                   1095:         sd*     at scsibus? target ? lun ?      # SCSI disk drives
                   1096:         cd*     at scsibus? target ? lun ?      # SCSI CD-ROM drives
1.1       mspo     1097: 
                   1098: 
1.28      gdt      1099: NetBSD as a domU in a VPS
                   1100: =========================
                   1101: 
                   1102: The bulk of the HOWTO is about using NetBSD as a dom0 on your own
                   1103: hardware.  This section explains how to deal with Xen in a domU as a
                   1104: virtual private server where you do not control or have access to the
1.70      gdt      1105: dom0.  This is not intended to be an exhaustive list of VPS providers;
                   1106: only a few are mentioned that specifically support NetBSD.
1.28      gdt      1107: 
1.52      gdt      1108: VPS operators provide varying degrees of access and mechanisms for
                   1109: configuration.  The big issue is usually how one controls which kernel
1.111     wiz      1110: is booted, because the kernel is nominally in the dom0 file system (to
                   1111: which VPS users do not normally have access).  A second issue is how
1.70      gdt      1112: to install NetBSD.
1.52      gdt      1113: A VPS user may want to compile a kernel for security updates, to run
                   1114: npf, run IPsec, or any other reason why someone would want to change
                   1115: their kernel.
                   1116: 
1.111     wiz      1117: One approach is to have an administrative interface to upload a kernel,
1.68      gdt      1118: or to select from a prepopulated list.  Other approaches are pygrub
1.59      gdt      1119: (deprecated) and pvgrub, which are ways to have a bootloader obtain a
1.111     wiz      1120: kernel from the domU file system.  This is closer to a regular physical
1.59      gdt      1121: computer, where someone who controls a machine can replace the kernel.
1.52      gdt      1122: 
1.74      gdt      1123: A second issue is multiple CPUs.  With NetBSD 6, domUs support
                   1124: multiple vcpus, and it is typical for VPS providers to enable multiple
                   1125: CPUs for NetBSD domUs.
                   1126: 
1.68      gdt      1127: pygrub
1.59      gdt      1128: -------
1.52      gdt      1129: 
1.111     wiz      1130: pygrub runs in the dom0 and looks into the domU file system.  This
                   1131: implies that the domU must have a kernel in a file system in a format
1.68      gdt      1132: known to pygrub.  As of 2014, pygrub seems to be of mostly historical
                   1133: interest.
1.52      gdt      1134: 
1.59      gdt      1135: pvgrub
                   1136: ------
                   1137: 
                   1138: pvgrub is a version of grub that uses PV operations instead of BIOS
                   1139: calls.  It is booted from the dom0 as the domU kernel, and then reads
1.111     wiz      1140: /grub/menu.lst and loads a kernel from the domU file system.
1.59      gdt      1141: 
1.70      gdt      1142: [Panix](http://www.panix.com/) lets users use pvgrub.  Panix reports
1.71      gdt      1143: that pvgrub works with FFsv2 with 16K/2K and 32K/4K block/frag sizes
                   1144: (and hence with defaults from "newfs -O 2").  See [Panix's pvgrub
1.70      gdt      1145: page](http://www.panix.com/v-colo/grub.html), which describes only
1.74      gdt      1146: Linux but should be updated to cover NetBSD :-).
1.70      gdt      1147: 
                   1148: [prgmr.com](http://prgmr.com/) also lets users with pvgrub to boot
                   1149: their own kernel.  See then [prgmr.com NetBSD
1.74      gdt      1150: HOWTO](http://wiki.prgmr.com/mediawiki/index.php/NetBSD_as_a_DomU)
                   1151: (which is in need of updating).
1.59      gdt      1152: 
1.70      gdt      1153: It appears that [grub's FFS
                   1154: code](http://xenbits.xensource.com/hg/xen-unstable.hg/file/bca284f67702/tools/libfsimage/ufs/fsys_ufs.c)
                   1155: does not support all aspects of modern FFS, but there are also reports
1.72      gdt      1156: that FFSv2 works fine.  At prgmr, typically one has an ext2 or FAT
1.70      gdt      1157: partition for the kernel with the intent that grub can understand it,
                   1158: which leads to /netbsd not being the actual kernel.  One must remember
1.111     wiz      1159: to update the special boot partition.
1.59      gdt      1160: 
                   1161: Amazon
                   1162: ------
                   1163: 
1.95      gdt      1164: See the [Amazon EC2 page](../amazon_ec2/).
1.44      gdt      1165: 
                   1166: Using npf
                   1167: ---------
                   1168: 
1.81      gdt      1169: In standard kernels, npf is a module, and thus cannot be loaded in a
1.44      gdt      1170: DOMU kernel.
                   1171: 
1.95      gdt      1172: TODO: Explain how to compile npf into a custom kernel, answering (but
                   1173: note that the problem was caused by not booting the right kernel)
                   1174: [this email to
1.140   ! khorben  1175: netbsd-users](https://mail-index.netbsd.org/netbsd-users/2014/12/26/msg015576.html).
1.65      gdt      1176: 
                   1177: TODO items for improving NetBSD/xen
                   1178: ===================================
                   1179: 
1.93      gdt      1180: * Make the NetBSD dom0 kernel work with SMP.
                   1181: * Test the Xen 4.5 packages adequately to be able to recommend them as
                   1182:   the standard approach.
                   1183: * Get PCI passthrough working on Xen 4.5
1.65      gdt      1184: * Get pvgrub into pkgsrc, either via xentools or separately.
                   1185: * grub
1.70      gdt      1186:   * Check/add support to pkgsrc grub2 for UFS2 and arbitrary
1.66      gdt      1187:     fragsize/blocksize (UFS2 support may be present; the point is to
1.111     wiz      1188:     make it so that with any UFS1/UFS2 file system setup that works
1.66      gdt      1189:     with NetBSD grub will also work).
1.140   ! khorben  1190:     See [pkg/40258](https://gnats.netbsd.org/40258).
1.65      gdt      1191:   * Push patches upstream.
                   1192:   * Get UFS2 patches into pvgrub.
                   1193: * Add support for PV ops to a version of /boot, and make it usable as
                   1194:   a kernel in Xen, similar to pvgrub.
1.93      gdt      1195: * Solve somehow the issue with modules for GENERIC not being loadable
                   1196:   in a Xen dom0 or domU kernel.
                   1197: 
                   1198: Random pointers
                   1199: ===============
                   1200: 
1.117     gdt      1201: This section contains links from elsewhere not yet integrated into the
                   1202: HOWTO, and other guides.
1.93      gdt      1203: 
                   1204: * http://www.lumbercartel.ca/library/xen/
                   1205: * http://pbraun.nethence.com/doc/sysutils/xen_netbsd_dom0.html
1.117     gdt      1206: * https://gmplib.org/~tege/xen.html

CVSweb for NetBSD wikisrc <wikimaster@NetBSD.org> software: FreeBSD-CVSweb