wikisrc/ports/xen/howto.mdwn - view

File: [NetBSD Developer Wiki] / wikisrc / ports / xen / howto.mdwn
Revision 1.183: download - view: text, annotated - select for diffs
Sun Nov 15 14:52:16 2020 UTC (2 months ago) by gdt
Branches: MAIN
CVS tags: HEAD

explain netbsd versions

1: [[!meta title="Xen HowTo"]] 2: 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. 10: 11: This HOWTO presumes a basic familiarity with the Xen system 12: architecture, with installing NetBSD on amd64 hardware, and with 13: installing software from pkgsrc. See also the [Xen 14: website](http://www.xenproject.org/). 15: 16: [[!toc]] 17: 18: # Overview 19: 20: The basic concept of Xen is that the hypervisor (xenkernel) runs on 21: the hardware, and runs a privileged domain ("dom0") that can access 22: disks/networking/etc. One then runs additonal unprivileged domains 23: (each a "domU"), presumably to do something useful. 24: 25: This HOWTO addresses how to run a NetBSD dom0 (and hence also build 26: xen itself). It also addresses how to run domUs in that environment, 27: and how to deal with having a domU in a Xen environment run by someone 28: else and/or not running NetBSD. 29: 30: ## Guest Styles 31: 32: Xen supports different styles of guests. 33: 34: [[!table data=""" 35: Style of guest |Supported by NetBSD 36: PV |Yes (dom0, domU) 37: HVM |Yes (domU) 38: PVHVM |current-only (domU) 39: PVH |current-only (domU, dom0 not yet) 40: """]] 41: 42: In Para-Virtualized (PV) mode, the guest OS does not attempt to access 43: hardware directly, but instead makes hypercalls to the hypervisor; PV 44: guests must be specifically coded for Xen. 45: See [PV](https://wiki.xen.org/wiki/Paravirtualization_(PV\)). 46: 47: In HVM mode, no guest modification is required; however, hardware 48: support is required, such as VT-x on Intel CPUs and SVM on AMD CPUs. 49: The dom0 runs qemu to emulate hardware. 50: 51: In PVHVM mode, the guest runs as HVM, but additionally can use PV 52: drivers for efficiency. 53: See [PV on HVM](https://wiki.xen.org/wiki/PV_on_HVM). 54: 55: There have been two PVH modes: original PVH and PVHv2. Original PVH 56: was based on PV mode and is no longer relevant at all. PVHv2 is 57: basically lightweight HVM with PV drivers. A critical feature of it 58: is that qemu is not needed; the hypervisor can do the emulation that 59: is required. Thus, a dom0 can be PVHv2. 60: The source code uses PVH and config files use pvh; this refers to PVHv2. 61: See [PVH(v2)](https://wiki.xenproject.org/wiki/PVH_(v2\)_Domu). 62: 63: At system boot, the dom0 kernel is loaded as a module with Xen as the kernel. 64: The dom0 can start one or more domUs. (Booting is explained in detail 65: in the dom0 section.) 66: 67: ## CPU Architecture 68: 69: Xen runs on x86_64 hardware (the NetBSD amd64 port). 70: 71: There is a concept of Xen running on ARM, but there are no reports of this working with NetBSD. 72: 73: The dom0 system should be amd64. (Instructions for i386PAE dom0 have been removed from the HOWTO.) 74: 75: The domU can be i386PAE or amd64. 76: i386PAE at one point was considered as [faster](https://lists.xen.org/archives/html/xen-devel/2012-07/msg00085.html) than amd64. 77: 78: ## Xen Versions 79: 80: In NetBSD, Xen is provided in pkgsrc, via matching pairs of packages 81: xenkernel and xentools. We will refer only to the kernel versions, 82: but note that both packages must be installed together and must have 83: matching versions. 84: 85: Versions available in pkgsrc: 86: 87: [[!table data=""" 88: Xen Version |Package Name |Xen CPU Support |xm? |EOL'ed By Upstream 89: 4.11 |xenkernel411 |x86_64 | |No 90: 4.13 |xenkernel413 |x86_64 | |No 91: """]] 92: 93: See also the [Xen Security Advisory page](http://xenbits.xen.org/xsa/). 94: 95: Multiprocessor (SMP) support in NetBSD differs depending on the domain: 96: 97: [[!table data=""" 98: Domain |Supports SMP 99: dom0 |No 100: domU |Yes 101: """]] 102: 103: Note: NetBSD support is called XEN3. However, it does support Xen 4, 104: because the hypercall interface has remained identical. 105: 106: Older Xen had a python-based management tool called xm, now replaced 107: by xl. 108: 109: ## NetBSD versions 110: 111: Xen has been supported in NetBSD for a long time, at least since 2005. 112: Initially Xen was PV only. 113: 114: NetBSD 8 and up support PV and HVM modes. 115: 116: Support for PVHVM and PVH is available only in NetBSD-current. 117: 118: # Creating a dom0 119: 120: In order to install a NetBSD as a dom0, one must first install a normal 121: NetBSD system, and then pivot the install to a dom0 install by changing 122: the kernel and boot configuration. 123: 124: In 2018-05, trouble booting a dom0 was reported with 256M of RAM: with 125: 512M it worked reliably. This does not make sense, but if you see 126: "not ELF" after Xen boots, try increasing dom0 RAM. 127: 128: ## Installation of NetBSD 129: 130: [Install NetBSD/amd64](/guide/inst/) 131: just as you would if you were not using Xen. 132: 133: ## Installation of Xen 134: 135: We will consider that you chose to use Xen 4.13, with NetBSD/amd64 as 136: dom0. In the dom0, install xenkernel48 and xentools48 from pkgsrc. 137: 138: Once this is done, install the Xen kernel itself: 139: 140: [[!template id=programlisting text=""" 141: # cp /usr/pkg/xen48-kernel/xen.gz / 142: """]] 143: 144: Then, place a NetBSD XEN3_DOM0 kernel in the `/` directory. Such kernel 145: can either be compiled manually, or downloaded from the NetBSD FTP, for 146: example at: 147: 148: [[!template id=programlisting text=""" 149: ftp.netbsd.org/pub/NetBSD/NetBSD-8.0/amd64/binary/kernel/netbsd-XEN3_DOM0.gz 150: """]] 151: 152: Add a line to /boot.cfg to boot Xen: 153: 154: [[!template id=filecontent name="/boot.cfg" text=""" 155: menu=Xen:load /netbsd-XEN3_DOM0.gz console=pc;multiboot /xen.gz dom0_mem=512M 156: """]] 157: 158: This specifies that the dom0 should have 512MB of ram, leaving the rest 159: to be allocated for domUs. To use a serial console, use: 160: 161: [[!template id=filecontent name="/boot.cfg" text=""" 162: menu=Xen:load /netbsd-XEN3_DOM0.gz;multiboot /xen.gz dom0_mem=512M console=com1 com1=9600,8n1 163: """]] 164: 165: which will use the first serial port for Xen (which counts starting 166: from 1, unlike NetBSD which counts starting from 0), forcing 167: speed/parity. Because the NetBSD command line lacks a 168: "console=pc" argument, it will use the default "xencons" console device, 169: which directs the console I/O through Xen to the same console device Xen 170: itself uses (in this case, the serial port). 171: 172: In an attempt to add performance, one can also add `dom0_max_vcpus=1 dom0_vcpus_pin`, 173: to force only one vcpu to be provided (since NetBSD dom0 can't use 174: more) and to pin that vcpu to a physical CPU. Xen has 175: [many boot options](http://xenbits.xenproject.org/docs/4.13-testing/misc/xen-command-line.html), 176: and other than dom0 memory and max_vcpus, they are generally not 177: necessary. 178: 179: Copy the boot scripts into `/etc/rc.d`: 180: 181: [[!template id=programlisting text=""" 182: # cp /usr/pkg/share/examples/rc.d/xen* /etc/rc.d/ 183: """]] 184: 185: Enable `xencommons`: 186: 187: [[!template id=filecontent name="/etc/rc.conf" text=""" 188: xencommons=YES 189: """]] 190: 191: Now, reboot so that you are running a DOM0 kernel under Xen, rather 192: than GENERIC without Xen. 193: 194: TODO: Recommend for/against xen-watchdog. 195: 196: Once the reboot is done, use `xl` to inspect Xen's boot messages, 197: available resources, and running domains. For example: 198: 199: [[!template id=programlisting text=""" 200: # xl dmesg 201: ... xen's boot info ... 202: # xl info 203: ... available memory, etc ... 204: # xl list 205: Name Id Mem(MB) CPU State Time(s) Console 206: Domain-0 0 64 0 r---- 58.1 207: """]] 208: 209: Xen logs will be in /var/log/xen. 210: 211: ### Issues with xencommons 212: 213: `xencommons` starts `xenstored`, which stores data on behalf of dom0 and 214: domUs. It does not currently work to stop and start xenstored. 215: Certainly all domUs should be shutdown first, following the sort order 216: of the rc.d scripts. However, the dom0 sets up state with xenstored, 217: and is not notified when xenstored exits, leading to not recreating 218: the state when the new xenstored starts. Until there's a mechanism to 219: make this work, one should not expect to be able to restart xenstored 220: (and thus xencommons). There is currently no reason to expect that 221: this will get fixed any time soon. 222: 223: ## anita (for testing NetBSD) 224: 225: With the setup so far, one should be able to run 226: anita (see pkgsrc/misc/py-anita) to test NetBSD releases, by doing (as 227: root, because anita must create a domU): 228: 229: [[!template id=programlisting text=""" 230: anita --vmm=xl test file:///usr/obj/i386/ 231: """]] 232: 233: ## Xen-specific NetBSD issues 234: 235: There are (at least) two additional things different about NetBSD as a 236: dom0 kernel compared to hardware. 237: 238: One is that the module ABI is different because some of the #defines 239: change, so one must build modules for Xen. As of netbsd-7, the build 240: system does this automatically. 241: 242: The other difference is that XEN3_DOM0 does not have exactly the same 243: options as GENERIC. While it is debatable whether or not this is a 244: bug, users should be aware of this and can simply add missing config 245: items if desired. 246: 247: ## Updating NetBSD in a dom0 248: 249: This is just like updating NetBSD on bare hardware, assuming the new 250: version supports the version of Xen you are running. Generally, one 251: replaces the kernel and reboots, and then overlays userland binaries 252: and adjusts `/etc`. 253: 254: Note that one must update both the non-Xen kernel typically used for 255: rescue purposes and the DOM0 kernel used with Xen. 256: 257: ## Converting from grub to /boot 258: 259: These instructions were used to convert a system from 260: grub to /boot. The system was originally installed in February of 261: 2006 with a RAID1 setup and grub to boot Xen 2, and has been updated 262: over time. Before these commands, it was running NetBSD 6 i386, Xen 263: 4.1 and grub, much like the message linked earlier in the grub 264: section. 265: 266: [[!template id=programlisting text=""" 267: # Install MBR bootblocks on both disks. 268: fdisk -i /dev/rwd0d 269: fdisk -i /dev/rwd1d 270: # Install NetBSD primary boot loader (/ is FFSv1) into RAID1 components. 271: installboot -v /dev/rwd0d /usr/mdec/bootxx_ffsv1 272: installboot -v /dev/rwd1d /usr/mdec/bootxx_ffsv1 273: # Install secondary boot loader 274: cp -p /usr/mdec/boot / 275: # Create boot.cfg following earlier guidance: 276: menu=Xen:load /netbsd-XEN3PAE_DOM0.gz console=pc;multiboot /xen.gz dom0_mem=512M 277: menu=Xen.ok:load /netbsd-XEN3PAE_DOM0.ok.gz console=pc;multiboot /xen.ok.gz dom0_mem=512M 278: menu=GENERIC:boot 279: menu=GENERIC single-user:boot -s 280: menu=GENERIC.ok:boot netbsd.ok 281: menu=GENERIC.ok single-user:boot netbsd.ok -s 282: menu=Drop to boot prompt:prompt 283: default=1 284: timeout=30 285: """]] 286: 287: ## Upgrading Xen versions 288: 289: Minor version upgrades are trivial. Just rebuild/replace the 290: xenkernel version and copy the new xen.gz to `/` (where `/boot.cfg` 291: references it), and reboot. 292: 293: #Unprivileged domains (domU) 294: 295: This section describes general concepts about domUs. It does not 296: address specific domU operating systems or how to install them. The 297: config files for domUs are typically in `/usr/pkg/etc/xen`, and are 298: typically named so that the file name, domU name and the domU's host 299: name match. 300: 301: The domU is provided with CPU and memory by Xen, configured by the 302: dom0. The domU is provided with disk and network by the dom0, 303: mediated by Xen, and configured in the dom0. 304: 305: Entropy in domUs can be an issue; physical disks and network are on 306: the dom0. NetBSD's /dev/random system works, but is often challenged. 307: 308: ## Config files 309: 310: See /usr/pkg/share/examples/xen/xlexample* 311: for a small number of well-commented examples, mostly for running 312: GNU/Linux. 313: 314: The following is an example minimal domain configuration file. The domU 315: serves as a network file server. 316: 317: [[!template id=filecontent name="/usr/pkg/etc/xen/foo" text=""" 318: name = "domU-id" 319: kernel = "/netbsd-XEN3PAE_DOMU-i386-foo.gz" 320: memory = 1024 321: vif = [ 'mac=aa:00:00:d1:00:09,bridge=bridge0' ] 322: disk = [ 'file:/n0/xen/foo-wd0,0x0,w', 323: 'file:/n0/xen/foo-wd1,0x1,w' ] 324: """]] 325: 326: The domain will have name given in the `name` setting. The kernel has the 327: host/domU name in it, so that on the dom0 one can update the various 328: domUs independently. The `vif` line causes an interface to be provided, 329: with a specific mac address (do not reuse MAC addresses!), in bridge 330: mode. Two disks are provided, and they are both writable; the bits 331: are stored in files and Xen attaches them to a vnd(4) device in the 332: dom0 on domain creation. The system treats xbd0 as the boot device 333: without needing explicit configuration. 334: 335: By convention, domain config files are kept in `/usr/pkg/etc/xen`. Note 336: that "xl create" takes the name of a config file, while other commands 337: take the name of a domain. 338: 339: Examples of commands: 340: 341: [[!template id=programlisting text=""" 342: xl create /usr/pkg/etc/xen/foo 343: xl console domU-id 344: xl create -c /usr/pkg/etc/xen/foo 345: xl shutdown domU-id 346: xl list 347: """]] 348: 349: Typing `^]` will exit the console session. Shutting down a domain is 350: equivalent to pushing the power button; a NetBSD domU will receive a 351: power-press event and do a clean shutdown. Shutting down the dom0 352: will trigger controlled shutdowns of all configured domUs. 353: 354: ## CPU and memory 355: 356: A domain is provided with some number of vcpus, up to the number 357: of CPUs seen by the hypervisor. For a domU, it is controlled 358: from the config file by the "vcpus = N" directive. 359: 360: A domain is provided with memory; this is controlled in the config 361: file by "memory = N" (in megabytes). In the straightforward case, the 362: sum of the the memory allocated to the dom0 and all domUs must be less 363: than the available memory. 364: 365: Xen also provides a "balloon" driver, which can be used to let domains 366: use more memory temporarily. 367: 368: ## Virtual disks 369: 370: In domU config files, the disks are defined as a sequence of 3-tuples: 371: 372: * The first element is "method:/path/to/disk". Common methods are 373: "file:" for a file-backed vnd, and "phy:" for something that is already 374: a device, such as an LVM logical volume. 375: 376: * The second element is an artifact of how virtual disks are passed to 377: Linux, and a source of confusion with NetBSD Xen usage. Linux domUs 378: are given a device name to associate with the disk, and values like 379: "hda1" or "sda1" are common. In a NetBSD domU, the first disk appears 380: as xbd0, the second as xbd1, and so on. However, xl demands a 381: second argument. The name given is converted to a major/minor by 382: calling stat(2) on the name in /dev and this is passed to the domU. 383: In the general case, the dom0 and domU can be different operating 384: systems, and it is an unwarranted assumption that they have consistent 385: numbering in /dev, or even that the dom0 OS has a /dev. With NetBSD 386: as both dom0 and domU, using values of 0x0 for the first disk and 0x1 387: for the second works fine and avoids this issue. For a GNU/Linux 388: guest, one can create /dev/hda1 in /dev, or to pass 0x301 for 389: /dev/hda1. 390: 391: * The third element is "w" for writable disks, and "r" for read-only 392: disks. 393: 394: Example: 395: [[!template id=filecontent name="/usr/pkg/etc/xen/foo" text=""" 396: disk = [ 'file:/n0/xen/foo-wd0,0x0,w' ] 397: """]] 398: 399: Note that NetBSD by default creates only vnd[0123]. If you need more 400: than 4 total virtual disks at a time, run e.g. "./MAKEDEV vnd4" in the 401: dom0. 402: 403: Note that NetBSD by default creates only xbd[0123]. If you need more 404: virtual disks in a domU, run e.g. "./MAKEDEV xbd4" in the domU. 405: 406: Virtual Networking 407: ------------------ 408: 409: Xen provides virtual Ethernets, each of which connects the dom0 and a 410: domU. For each virtual network, there is an interface "xvifN.M" in 411: the dom0, and a matching interface xennetM (NetBSD name) in domU index N. 412: The interfaces behave as if there is an Ethernet with two 413: adapters connected. From this primitive, one can construct various 414: configurations. We focus on two common and useful cases for which 415: there are existing scripts: bridging and NAT. 416: 417: With bridging (in the example above), the domU perceives itself to be 418: on the same network as the dom0. For server virtualization, this is 419: usually best. Bridging is accomplished by creating a bridge(4) device 420: and adding the dom0's physical interface and the various xvifN.0 421: interfaces to the bridge. One specifies "bridge=bridge0" in the domU 422: config file. The bridge must be set up already in the dom0; an 423: example /etc/ifconfig.bridge0 is: 424: 425: [[!template id=filecontent name="/etc/ifconfig.bridge0" text=""" 426: create 427: up 428: !brconfig bridge0 add wm0 429: """]] 430: 431: With NAT, the domU perceives itself to be behind a NAT running on the 432: dom0. This is often appropriate when running Xen on a workstation. 433: TODO: NAT appears to be configured by "vif = [ '' ]". 434: 435: The MAC address specified is the one used for the interface in the new 436: domain. The interface in dom0 will use this address XOR'd with 437: 00:00:00:01:00:00. Random MAC addresses are assigned if not given. 438: 439: Starting domains automatically 440: ------------------------------ 441: 442: To start domains `domU-netbsd` and `domU-linux` at boot and shut them 443: down cleanly on dom0 shutdown, add the following in rc.conf: 444: 445: [[!template id=filecontent name="/etc/rc.conf" text=""" 446: xendomains="domU-netbsd domU-linux" 447: """]] 448: 449: # Creating a domU 450: 451: Creating domUs is almost entirely independent of operating system. We 452: have already presented the basics of config files. Note that you must 453: have already completed the dom0 setup so that "xl list" works. 454: 455: Creating a NetBSD PV domU 456: -------------------------- 457: 458: See the earlier config file, and adjust memory. Decide on how much 459: storage you will provide, and prepare it (file or LVM). 460: 461: While the kernel will be obtained from the dom0 file system, the same 462: file should be present in the domU as /netbsd so that tools like 463: savecore(8) can work. (This is helpful but not necessary.) 464: 465: The kernel must be specifically for Xen and for use as a domU. The 466: i386 and amd64 provide the following kernels: 467: 468: i386 XEN3PAE_DOMU 469: amd64 XEN3_DOMU 470: 471: This will boot NetBSD, but this is not that useful if the disk is 472: empty. One approach is to unpack sets onto the disk outside of xen 473: (by mounting it, just as you would prepare a physical disk for a 474: system you can't run the installer on). 475: 476: A second approach is to run an INSTALL kernel, which has a miniroot 477: and can load sets from the network. To do this, copy the INSTALL 478: kernel to / and change the kernel line in the config file to: 479: 480: kernel = "/home/bouyer/netbsd-INSTALL_XEN3_DOMU" 481: 482: Then, start the domain as "xl create -c configfile". 483: 484: Alternatively, if you want to install NetBSD/Xen with a CDROM image, the following 485: line should be used in the config file. 486: 487: disk = [ 'phy:/dev/wd0e,0x1,w', 'phy:/dev/cd0a,0x2,r' ] 488: 489: After booting the domain, the option to install via CDROM may be 490: selected. The CDROM device should be changed to `xbd1d`. 491: 492: Once done installing, "halt -p" the new domain (don't reboot or halt, 493: it would reload the INSTALL_XEN3_DOMU kernel even if you changed the 494: config file), switch the config file back to the XEN3_DOMU kernel, 495: and start the new domain again. Now it should be able to use "root on 496: xbd0a" and you should have a, functional NetBSD domU. 497: 498: TODO: check if this is still accurate. 499: When the new domain is booting you'll see some warnings about *wscons* 500: and the pseudo-terminals. These can be fixed by editing the files 501: `/etc/ttys` and `/etc/wscons.conf`. You must disable all terminals in 502: `/etc/ttys`, except *console*, like this: 503: 504: console "/usr/libexec/getty Pc" vt100 on secure 505: ttyE0 "/usr/libexec/getty Pc" vt220 off secure 506: ttyE1 "/usr/libexec/getty Pc" vt220 off secure 507: ttyE2 "/usr/libexec/getty Pc" vt220 off secure 508: ttyE3 "/usr/libexec/getty Pc" vt220 off secure 509: 510: Finally, all screens must be commented out from `/etc/wscons.conf`. 511: 512: It is also desirable to add 513: 514: powerd=YES 515: 516: in rc.conf. This way, the domain will be properly shut down if 517: `xm shutdown -R` or `xm shutdown -H` is used on the dom0. 518: 519: It is not strictly necessary to have a kernel (as /netbsd) in the domU 520: file system. However, various programs (e.g. netstat) will use that 521: kernel to look up symbols to read from kernel virtual memory. If 522: /netbsd is not the running kernel, those lookups will fail. (This is 523: not really a Xen-specific issue, but because the domU kernel is 524: obtained from the dom0, it is far more likely to be out of sync or 525: missing with Xen.) 526: 527: Creating a Linux domU 528: --------------------- 529: 530: Creating unprivileged Linux domains isn't much different from 531: unprivileged NetBSD domains, but there are some details to know. 532: 533: First, the second parameter passed to the disk declaration (the '0x1' in 534: the example below) 535: 536: disk = [ 'phy:/dev/wd0e,0x1,w' ] 537: 538: does matter to Linux. It wants a Linux device number here (e.g. 0x300 539: for hda). Linux builds device numbers as: (major \<\< 8 + minor). 540: So, hda1 which has major 3 and minor 1 on a Linux system will have 541: device number 0x301. Alternatively, devices names can be used (hda, 542: hdb, ...) as xentools has a table to map these names to devices 543: numbers. To export a partition to a Linux guest we can use: 544: 545: disk = [ 'phy:/dev/wd0e,0x300,w' ] 546: root = "/dev/hda1 ro" 547: 548: and it will appear as /dev/hda on the Linux system, and be used as root 549: partition. 550: 551: To install the Linux system on the partition to be exported to the 552: guest domain, the following method can be used: install 553: sysutils/e2fsprogs from pkgsrc. Use mke2fs to format the partition 554: that will be the root partition of your Linux domain, and mount it. 555: Then copy the files from a working Linux system, make adjustments in 556: `/etc` (fstab, network config). It should also be possible to extract 557: binary packages such as .rpm or .deb directly to the mounted partition 558: using the appropriate tool, possibly running under NetBSD's Linux 559: emulation. Once the file system has been populated, umount it. If 560: desirable, the file system can be converted to ext3 using tune2fs -j. 561: It should now be possible to boot the Linux guest domain, using one of 562: the vmlinuz-\*-xenU kernels available in the Xen binary distribution. 563: 564: To get the Linux console right, you need to add: 565: 566: extra = "xencons=tty1" 567: 568: to your configuration since not all Linux distributions auto-attach a 569: tty to the xen console. 570: 571: ## Creating a NetBSD HVM domU 572: 573: Use type='hmv', probably. Use a GENERIC kernel within the disk image. 574: 575: ## Creating a NetBSD PVH domU 576: 577: Use type='pvh'. 578: 579: \todo Explain where the kernel comes from. 580: 581: 582: Creating a Solaris domU 583: ----------------------- 584: 585: See possibly outdated 586: [Solaris domU instructions](/ports/xen/howto-solaris/). 587: 588: 589: PCI passthrough: Using PCI devices in guest domains 590: --------------------------------------------------- 591: 592: NB: PCI passthrough only works on some Xen versions and as of 2020 it 593: is not clear that it works on any version in pkgsrc. Reports 594: confirming or denying this notion should be sent to port-xen@. 595: 596: The dom0 can give other domains access to selected PCI 597: devices. This can allow, for example, a non-privileged domain to have 598: access to a physical network interface or disk controller. However, 599: keep in mind that giving a domain access to a PCI device most likely 600: will give the domain read/write access to the whole physical memory, 601: as PCs don't have an IOMMU to restrict memory access to DMA-capable 602: device. Also, it's not possible to export ISA devices to non-dom0 603: domains, which means that the primary VGA adapter can't be exported. 604: A guest domain trying to access the VGA registers will panic. 605: 606: If the dom0 is NetBSD, it has to be running Xen 3.1, as support has 607: not been ported to later versions at this time. 608: 609: For a PCI device to be exported to a domU, is has to be attached to 610: the "pciback" driver in dom0. Devices passed to the dom0 via the 611: pciback.hide boot parameter will attach to "pciback" instead of the 612: usual driver. The list of devices is specified as "(bus:dev.func)", 613: where bus and dev are 2-digit hexadecimal numbers, and func a 614: single-digit number: 615: 616: pciback.hide=(00:0a.0)(00:06.0) 617: 618: pciback devices should show up in the dom0's boot messages, and the 619: devices should be listed in the `/kern/xen/pci` directory. 620: 621: PCI devices to be exported to a domU are listed in the "pci" array of 622: the domU's config file, with the format "0000:bus:dev.func". 623: 624: pci = [ '0000:00:06.0', '0000:00:0a.0' ] 625: 626: In the domU an "xpci" device will show up, to which one or more pci 627: buses will attach. Then the PCI drivers will attach to PCI buses as 628: usual. Note that the default NetBSD DOMU kernels do not have "xpci" 629: or any PCI drivers built in by default; you have to build your own 630: kernel to use PCI devices in a domU. Here's a kernel config example; 631: note that only the "xpci" lines are unusual. 632: 633: include "arch/i386/conf/XEN3_DOMU" 634: 635: # Add support for PCI buses to the XEN3_DOMU kernel 636: xpci* at xenbus ? 637: pci* at xpci ? 638: 639: # PCI USB controllers 640: uhci* at pci? dev ? function ? # Universal Host Controller (Intel) 641: 642: # USB bus support 643: usb* at uhci? 644: 645: # USB Hubs 646: uhub* at usb? 647: uhub* at uhub? port ? configuration ? interface ? 648: 649: # USB Mass Storage 650: umass* at uhub? port ? configuration ? interface ? 651: wd* at umass? 652: # SCSI controllers 653: ahc* at pci? dev ? function ? # Adaptec [23]94x, aic78x0 SCSI 654: 655: # SCSI bus support (for both ahc and umass) 656: scsibus* at scsi? 657: 658: # SCSI devices 659: sd* at scsibus? target ? lun ? # SCSI disk drives 660: cd* at scsibus? target ? lun ? # SCSI CD-ROM drives 661: 662: 663: # Specific Issues 664: 665: ## domU 666: 667: [NetBSD 5 is known to panic.](http://mail-index.netbsd.org/port-xen/2018/04/17/msg009181.html) 668: (However, NetBSD 5 systems should be updated to a supported version.) 669: 670: # NetBSD as a domU in a VPS 671: 672: The bulk of the HOWTO is about using NetBSD as a dom0 on your own 673: hardware. This section explains how to deal with Xen in a domU as a 674: virtual private server where you do not control or have access to the 675: dom0. This is not intended to be an exhaustive list of VPS providers; 676: only a few are mentioned that specifically support NetBSD. 677: 678: VPS operators provide varying degrees of access and mechanisms for 679: configuration. The big issue is usually how one controls which kernel 680: is booted, because the kernel is nominally in the dom0 file system (to 681: which VPS users do not normally have access). A second issue is how 682: to install NetBSD. 683: A VPS user may want to compile a kernel for security updates, to run 684: npf, run IPsec, or any other reason why someone would want to change 685: their kernel. 686: 687: One approach is to have an administrative interface to upload a kernel, 688: or to select from a prepopulated list. Other approaches are pygrub 689: (deprecated) and pvgrub, which are ways to have a bootloader obtain a 690: kernel from the domU file system. This is closer to a regular physical 691: computer, where someone who controls a machine can replace the kernel. 692: 693: A second issue is multiple CPUs. With NetBSD 6, domUs support 694: multiple vcpus, and it is typical for VPS providers to enable multiple 695: CPUs for NetBSD domUs. 696: 697: ## Complexities due to Xen changes 698: 699: Xen has many security advisories and people running Xen systems make 700: different choices. 701: 702: ### stub domains 703: 704: Some (Linux only?) dom0 systems use something called "stub domains" to 705: isolate qemu from the dom0 system, as a security and reliabilty 706: mechanism when running HVM domUs. Somehow, NetBSD's GENERIC kernel 707: ends up using PIO for disks rather than DMA. Of course, all of this 708: is emulated, but emulated PIO is unusably slow. This problem is not 709: currently understood. 710: 711: ### Grant tables 712: 713: There are multiple versions of using grant tables, and some security 714: advisories have suggested disabling some versions. Some versions of 715: NetBSD apparently only use specific versions and this can lead to 716: "NetBSD current doesn't run on hosting provider X" situations. 717: 718: \todo Explain better. 719: 720: ## Boot methods 721: 722: ### pvgrub 723: 724: pvgrub is a version of grub that uses PV operations instead of BIOS 725: calls. It is booted from the dom0 as the domU kernel, and then reads 726: /grub/menu.lst and loads a kernel from the domU file system. 727: 728: [Panix](http://www.panix.com/) lets users use pvgrub. Panix reports 729: that pvgrub works with FFsv2 with 16K/2K and 32K/4K block/frag sizes 730: (and hence with defaults from "newfs -O 2"). See [Panix's pvgrub 731: page](http://www.panix.com/v-colo/grub.html), which describes only 732: Linux but should be updated to cover NetBSD :-). 733: 734: [prgmr.com](http://prgmr.com/) also lets users with pvgrub to boot 735: their own kernel. See then [prgmr.com NetBSD 736: HOWTO](http://wiki.prgmr.com/mediawiki/index.php/NetBSD_as_a_DomU) 737: (which is in need of updating). 738: 739: It appears that [grub's FFS 740: code](http://xenbits.xensource.com/hg/xen-unstable.hg/file/bca284f67702/tools/libfsimage/ufs/fsys_ufs.c) 741: does not support all aspects of modern FFS, but there are also reports 742: that FFSv2 works fine. At prgmr, typically one has an ext2 or FAT 743: partition for the kernel with the intent that grub can understand it, 744: which leads to /netbsd not being the actual kernel. One must remember 745: to update the special boot partition. 746: 747: ### pygrub 748: 749: pygrub runs in the dom0 and looks into the domU file system. This 750: implies that the domU must have a kernel in a file system in a format 751: known to pygrub. 752: 753: pygrub doesn't seem to work to load Linux images under NetBSD dom0, 754: and is inherently less secure than pvgrub due to running inside dom0. For both these 755: reasons, pygrub should not be used, and is only still present so that 756: historical DomU images using it still work. 757: 758: As of 2014, pygrub seems to be of mostly historical 759: interest. New DomUs should use pvgrub. 760: 761: ## Specific Providers 762: 763: ### Amazon 764: 765: See the [Amazon EC2 page](/amazon_ec2/).