Development Guide: build HVM IO domain on xVM

This document describes the instructions to build HVM IO domain. The hypervisor is based on xvm-3.3 and domain 0 based on onnv_121.

1. Upgrade domain 0

1) Install Nevada build 121 onto the system which supports Intel virtualization technology for direct IO. Make sure Intel vt-d is enabled in BIOS setting.

2) Build the BFU archive from onnv_121 with patches.

$ hg clone ssh://[e-mail]/hg/onnv/onnv-gate

$ cd onnv-gate

$ hg update -C onnv_121

$ hg qinit

$ hg qimport pci-device-reservation

$ hg qpush

$ hg qimport xen-hcall-extension

$ hg qpush

Download closed binary archives from http://dlc.sun.com/osol/on/downloads/b121/. Unpatch the tar ball under onnv-gate.

$ bzcat on-closed-bins-nd.i386.tar.bz2 | tar xf -

$ bzcat on-closed-bins.i386.tar.bz2 | tar xf -

Setup env script and run nightly

$ cp usr/src/tools/env/opensolaris.sh ./

$ edit opensolaris.sh

$ nightly opensolaris.sh

3) BFU the system installed by step 1 with the archive built during step 2.

2. Build xvm-3.3 packages

1) Pull a xvm-3.3 code base under local directory

$ wget http://dlc.sun.com/osol/on/downloads/b121/xvm-src.tar.bz2

Or,

Pull it from ssh://[e-mail]/hg/xen-gate/xvm-3.3+(sub-gates)

2) import the patch for xen.hg and qemu.hg

$ cd xen.hg

$ hg qpush -a

$ hg qimport python-pci-aug27

$ hg qpush

$ cd qemu.hg

$ hg qpush -a

$ hg qimport pci-passthrough-aug27

$ hg qpush

3) build the packages

$ export XVM_WS=`pwd`

$ ./sunos.hg/bin/build-all full

3. Upgrade xvm-3.3 packages

# svcadm disable xvm/domains xvm/console xvm/xend xvm/store xvm/virtd

# pkgrm SUNWlibvirt SUNWlibvirtr SUNWurlgrabber SUNWvdisk SUNWvirtinst SUNWxvmdomr SUNWxvmdomu SUNWxvmh SUNWxvmhvm SUNWxvmr SUNWxvmu

# pkgadd -d . SUNWlibvirt SUNWlibvirtr SUNWurlgrabber SUNWvdisk SUNWvirtinst SUNWxvmdomr SUNWxvmdomu SUNWxvmh SUNWxvmhvm SUNWxvmr SUNWxvmu

# svcadm enable xvm/domains xvm/console xvm/xend xvm/store xvm/virtd

# reboot

4. Enable xend to access pci device configure space

# svccfg -s xend setprop start/privileges = all

# svcadm refresh xend

# svcadm restart xend

5. Configure grub boot menu

# edit [/rpool]/boot/grub/menu.lst

title Solaris xVM

findroot (rootfs0,0,a)

kernel$ /boot/$ISADIR/xen.gz iommu=1

module$ /platform/i86xpv/kernel/$ISADIR/unix /platform/i86xpv/kernel/$ISADIR/unix -B pci-reserve="[1:0:0],[1:0:1],[3:0:0]"

module$ /platform/i86pc/$ISADIR/boot_archive

where, "iommu=1" enables the hypervisor to support Intel vt-d; "-B pci-reserve="[1:0:0],[1:0:1],[3:0:0]"" asks domain 0 to reserve pci devices for pass-through purpose.

6. pass through a pci device to HVM guest

1) below is a sample xm configure file for pci device pass-through

#### start of the configure file, pt-sample.py ####

import os, re

arch = os.uname()[4]

if re.search('64', arch):

arch_libdir = 'lib64'

else:

arch_libdir = 'lib'

kernel = "/usr/lib/xen/boot/hvmloader"

builder='hvm'

memory = 512

shadow_memory = 8

name = "r48"

vcpus=2

pci = [ '01:00.0' ]

disk = [ 'file:/export/home/allen/iodomain/images/ia32e_rhel4u8.img,hdc,w' ]

device_model = '/usr/' + arch_libdir + '/xen/bin/qemu-dm'

boot='c'

vnc=1

vnclisten="0.0.0.0"

vncconsole=1

vncpasswd=''

nographic=0

stdvga=0

serial='null'

on_poweroff = 'destroy'

on_reboot = 'restart'

on_crash = 'preserve'

#### end of the configure file, pt-sample.py ####

where, "pci = [ '01:00.0' ]" tells xm to pass through pci device with bdf [1,0,0] to the guest.

2) create the hvm guest domain

# xm create -c pt-sample.py

You will find the pass-through device, when you logon the domain.