nixpkgs/nixos/modules/virtualisation/qemu-vm.nix

518 lines
17 KiB
Nix
Raw Normal View History

# This module creates a virtual machine from the NixOS configuration.
# Building the `config.system.build.vm' attribute gives you a command
# that starts a KVM/QEMU VM running the NixOS configuration defined in
# `config'. The Nix store is shared read-only with the host, which
# makes (re)building VMs very efficient. However, it also means you
# can't reconfigure the guest inside the guest - you need to rebuild
# the VM in the host. On the other hand, the root filesystem is a
# read/writable disk image persistent across VM reboots.
{ config, lib, pkgs, ... }:
with lib;
let
vmName =
if config.networking.hostName == ""
then "noname"
else config.networking.hostName;
cfg = config.virtualisation;
qemuGraphics = if cfg.graphics then "" else "-nographic";
kernelConsole = if cfg.graphics then "" else "console=ttyS0";
ttys = [ "tty1" "tty2" "tty3" "tty4" "tty5" "tty6" ];
# Shell script to start the VM.
startVM =
''
#! ${pkgs.stdenv.shell}
NIX_DISK_IMAGE=$(readlink -f ''${NIX_DISK_IMAGE:-${config.virtualisation.diskImage}})
if ! test -e "$NIX_DISK_IMAGE"; then
${pkgs.qemu_kvm}/bin/qemu-img create -f qcow2 "$NIX_DISK_IMAGE" \
${toString config.virtualisation.diskSize}M || exit 1
fi
# Create a directory for storing temporary data of the running VM.
if [ -z "$TMPDIR" -o -z "$USE_TMPDIR" ]; then
TMPDIR=$(mktemp -d nix-vm.XXXXXXXXXX --tmpdir)
fi
# Create a directory for exchanging data with the VM.
mkdir -p $TMPDIR/xchg
${if cfg.useBootLoader then ''
# Create a writable copy/snapshot of the boot disk
# A writable boot disk can be booted from automatically
${pkgs.qemu_kvm}/bin/qemu-img create -f qcow2 -b ${bootDisk}/disk.img $TMPDIR/disk.img || exit 1
${if cfg.useEFIBoot then ''
# VM needs a writable flash BIOS
cp ${bootDisk}/bios.bin $TMPDIR || exit 1
chmod 0644 $TMPDIR/bios.bin || exit 1
'' else ''
''}
'' else ''
''}
cd $TMPDIR
idx=2
extraDisks=""
${flip concatMapStrings cfg.emptyDiskImages (size: ''
${pkgs.qemu_kvm}/bin/qemu-img create -f qcow2 "empty$idx.qcow2" "${toString size}M"
extraDisks="$extraDisks -drive index=$idx,file=$(pwd)/empty$idx.qcow2,if=${cfg.qemu.diskInterface},werror=report"
idx=$((idx + 1))
'')}
# Start QEMU.
exec ${pkgs.qemu_kvm}/bin/qemu-kvm \
-name ${vmName} \
-m ${toString config.virtualisation.memorySize} \
${optionalString (pkgs.stdenv.system == "x86_64-linux") "-cpu kvm64"} \
${concatStringsSep " " config.virtualisation.qemu.networkingOptions} \
-virtfs local,path=/nix/store,security_model=none,mount_tag=store \
-virtfs local,path=$TMPDIR/xchg,security_model=none,mount_tag=xchg \
-virtfs local,path=''${SHARED_DIR:-$TMPDIR/xchg},security_model=none,mount_tag=shared \
${if cfg.useBootLoader then ''
-drive index=0,id=drive1,file=$NIX_DISK_IMAGE,if=${cfg.qemu.diskInterface},cache=none,werror=report \
-drive index=1,id=drive2,file=$TMPDIR/disk.img,media=disk \
${if cfg.useEFIBoot then ''
-pflash $TMPDIR/bios.bin \
'' else ''
''}
'' else ''
-drive index=0,id=drive1,file=$NIX_DISK_IMAGE,if=${cfg.qemu.diskInterface},cache=none,werror=report \
-kernel ${config.system.build.toplevel}/kernel \
-initrd ${config.system.build.toplevel}/initrd \
-append "$(cat ${config.system.build.toplevel}/kernel-params) init=${config.system.build.toplevel}/init regInfo=${regInfo} ${kernelConsole} $QEMU_KERNEL_PARAMS" \
''} \
$extraDisks \
${qemuGraphics} \
${toString config.virtualisation.qemu.options} \
$QEMU_OPTS \
$@
'';
regInfo = pkgs.runCommand "reginfo"
{ exportReferencesGraph =
map (x: [("closure-" + baseNameOf x) x]) config.virtualisation.pathsInNixDB;
buildInputs = [ pkgs.perl ];
preferLocalBuild = true;
}
''
printRegistration=1 perl ${pkgs.pathsFromGraph} closure-* > $out
'';
# Generate a hard disk image containing a /boot partition and GRUB
# in the MBR. Used when the `useBootLoader' option is set.
bootDisk =
pkgs.vmTools.runInLinuxVM (
pkgs.runCommand "nixos-boot-disk"
{ preVM =
''
mkdir $out
diskImage=$out/disk.img
bootFlash=$out/bios.bin
${pkgs.qemu_kvm}/bin/qemu-img create -f qcow2 $diskImage "40M"
${if cfg.useEFIBoot then ''
cp ${pkgs.OVMF-CSM}/FV/OVMF.fd $bootFlash
chmod 0644 $bootFlash
'' else ''
''}
'';
buildInputs = [ pkgs.utillinux ];
QEMU_OPTS = if cfg.useEFIBoot
then "-pflash $out/bios.bin -nographic -serial pty"
else "-nographic -serial pty";
}
''
# Create a /boot EFI partition with 40M
${pkgs.gptfdisk}/sbin/sgdisk -G /dev/vda
${pkgs.gptfdisk}/sbin/sgdisk -a 1 -n 1:34:2047 -c 1:"BIOS Boot Partition" -t 1:ef02 /dev/vda
${pkgs.gptfdisk}/sbin/sgdisk -a 512 -N 2 -c 2:"EFI System" -t 2:ef00 /dev/vda
${pkgs.gptfdisk}/sbin/sgdisk -A 1:set:1 /dev/vda
${pkgs.gptfdisk}/sbin/sgdisk -A 2:set:2 /dev/vda
${pkgs.gptfdisk}/sbin/sgdisk -h 2 /dev/vda
${pkgs.gptfdisk}/sbin/sgdisk -C /dev/vda
${pkgs.utillinux}/bin/sfdisk /dev/vda -A 2
. /sys/class/block/vda2/uevent
mknod /dev/vda2 b $MAJOR $MINOR
. /sys/class/block/vda/uevent
${pkgs.dosfstools}/bin/mkfs.fat -F16 /dev/vda2
export MTOOLS_SKIP_CHECK=1
${pkgs.mtools}/bin/mlabel -i /dev/vda2 ::boot
# Mount /boot; load necessary modules first.
${pkgs.module_init_tools}/sbin/insmod ${pkgs.linux}/lib/modules/*/kernel/fs/nls/nls_cp437.ko || true
${pkgs.module_init_tools}/sbin/insmod ${pkgs.linux}/lib/modules/*/kernel/fs/nls/nls_iso8859-1.ko || true
${pkgs.module_init_tools}/sbin/insmod ${pkgs.linux}/lib/modules/*/kernel/fs/fat/fat.ko || true
${pkgs.module_init_tools}/sbin/insmod ${pkgs.linux}/lib/modules/*/kernel/fs/fat/vfat.ko || true
${pkgs.module_init_tools}/sbin/insmod ${pkgs.linux}/lib/modules/*/kernel/fs/efivarfs/efivarfs.ko || true
mkdir /boot
mount /dev/vda2 /boot
# This is needed for GRUB 0.97, which doesn't know about virtio devices.
mkdir /boot/grub
echo '(hd0) /dev/vda' > /boot/grub/device.map
# Install GRUB and generate the GRUB boot menu.
touch /etc/NIXOS
mkdir -p /nix/var/nix/profiles
${config.system.build.toplevel}/bin/switch-to-configuration boot
umount /boot
'' # */
);
in
{
imports = [ ../profiles/qemu-guest.nix ];
options = {
virtualisation.memorySize =
mkOption {
default = 384;
description =
''
Memory size (M) of virtual machine.
'';
};
virtualisation.diskSize =
mkOption {
default = 512;
description =
''
Disk size (M) of virtual machine.
'';
};
virtualisation.diskImage =
mkOption {
default = "./${vmName}.qcow2";
description =
''
Path to the disk image containing the root filesystem.
The image will be created on startup if it does not
exist.
'';
};
virtualisation.bootDevice =
mkOption {
type = types.str;
example = "/dev/vda";
description =
''
The disk to be used for the root filesystem.
'';
};
virtualisation.emptyDiskImages =
mkOption {
default = [];
2013-09-04 13:12:07 +00:00
type = types.listOf types.int;
description =
''
Additional disk images to provide to the VM. The value is
a list of size in megabytes of each disk. These disks are
writeable by the VM.
'';
};
virtualisation.graphics =
mkOption {
default = true;
description =
''
Whether to run QEMU with a graphics window, or access
the guest computer serial port through the host tty.
'';
};
virtualisation.pathsInNixDB =
mkOption {
default = [];
description =
''
The list of paths whose closure is registered in the Nix
database in the VM. All other paths in the host Nix store
appear in the guest Nix store as well, but are considered
garbage (because they are not registered in the Nix
database in the guest).
'';
};
virtualisation.vlans =
mkOption {
default = [ 1 ];
example = [ 1 2 ];
description =
''
Virtual networks to which the VM is connected. Each
number <replaceable>N</replaceable> in this list causes
the VM to have a virtual Ethernet interface attached to a
separate virtual network on which it will be assigned IP
address
<literal>192.168.<replaceable>N</replaceable>.<replaceable>M</replaceable></literal>,
where <replaceable>M</replaceable> is the index of this VM
in the list of VMs.
'';
};
virtualisation.writableStore =
mkOption {
default = false;
description =
''
If enabled, the Nix store in the VM is made writable by
layering a unionfs-fuse/tmpfs filesystem on top of the host's Nix
store.
'';
};
virtualisation.writableStoreUseTmpfs =
mkOption {
default = true;
description =
''
Use a tmpfs for the writable store instead of writing to the VM's
own filesystem.
'';
};
networking.primaryIPAddress =
mkOption {
default = "";
internal = true;
description = "Primary IP address used in /etc/hosts.";
};
virtualisation.qemu = {
options =
mkOption {
type = types.listOf types.unspecified;
default = [];
example = [ "-vga std" ];
description = "Options passed to QEMU.";
};
networkingOptions =
mkOption {
default = [
"-net nic,vlan=0,model=virtio"
"-net user,vlan=0\${QEMU_NET_OPTS:+,$QEMU_NET_OPTS}"
];
type = types.listOf types.str;
description = ''
Networking-related command-line options that should be passed to qemu.
The default is to use userspace networking (slirp).
If you override this option, be advised to keep
''${QEMU_NET_OPTS:+,$QEMU_NET_OPTS} (as seen in the default)
to keep the default runtime behaviour.
'';
};
diskInterface =
mkOption {
default = "virtio";
example = "scsi";
type = types.str;
description = ''
The interface used for the virtual hard disks
(<literal>virtio</literal> or <literal>scsi</literal>).
'';
};
};
virtualisation.useBootLoader =
mkOption {
default = false;
description =
''
If enabled, the virtual machine will be booted using the
regular boot loader (i.e., GRUB 1 or 2). This allows
testing of the boot loader. If
disabled (the default), the VM directly boots the NixOS
kernel and initial ramdisk, bypassing the boot loader
altogether.
'';
};
virtualisation.useEFIBoot =
mkOption {
default = false;
description =
''
If enabled, the virtual machine will provide a EFI boot
manager.
useEFIBoot is ignored if useBootLoader == false.
'';
};
};
config = {
boot.loader.grub.device = mkVMOverride cfg.bootDevice;
boot.initrd.extraUtilsCommands =
''
# We need mke2fs in the initrd.
2015-03-29 00:15:41 +00:00
copy_bin_and_libs ${pkgs.e2fsprogs}/sbin/mke2fs
'';
boot.initrd.postDeviceCommands =
''
# If the disk image appears to be empty, run mke2fs to
# initialise.
FSTYPE=$(blkid -o value -s TYPE ${cfg.bootDevice} || true)
if test -z "$FSTYPE"; then
mke2fs -t ext4 ${cfg.bootDevice}
fi
'';
boot.initrd.postMountCommands =
''
# Mark this as a NixOS machine.
mkdir -p $targetRoot/etc
echo -n > $targetRoot/etc/NIXOS
# Fix the permissions on /tmp.
chmod 1777 $targetRoot/tmp
mkdir -p $targetRoot/boot
'';
# After booting, register the closure of the paths in
# `virtualisation.pathsInNixDB' in the Nix database in the VM. This
# allows Nix operations to work in the VM. The path to the
# registration file is passed through the kernel command line to
# allow `system.build.toplevel' to be included. (If we had a direct
# reference to ${regInfo} here, then we would get a cyclic
# dependency.)
boot.postBootCommands =
''
if [[ "$(cat /proc/cmdline)" =~ regInfo=([^ ]*) ]]; then
2013-10-28 15:28:04 +00:00
${config.nix.package}/bin/nix-store --load-db < ''${BASH_REMATCH[1]}
fi
'';
boot.initrd.availableKernelModules =
optional (cfg.qemu.diskInterface == "scsi") "sym53c8xx";
virtualisation.bootDevice =
mkDefault (if cfg.qemu.diskInterface == "scsi" then "/dev/sda" else "/dev/vda");
virtualisation.pathsInNixDB = [ config.system.build.toplevel ];
virtualisation.qemu.options = [ "-vga std" "-usbdevice tablet" ];
# Mount the host filesystem via 9P, and bind-mount the Nix store
# of the host into our own filesystem. We use mkVMOverride to
# allow this module to be applied to "normal" NixOS system
# configuration, where the regular value for the `fileSystems'
# attribute should be disregarded for the purpose of building a VM
# test image (since those filesystems don't exist in the VM).
fileSystems = mkVMOverride (
{ "/".device = cfg.bootDevice;
${if cfg.writableStore then "/nix/.ro-store" else "/nix/store"} =
{ device = "store";
fsType = "9p";
nixos/vm-tests: Remove msize mount option This seems to be the root cause of the random page allocation failures and @wizeman did a very good job on not only finding the root problem but also giving a detailed explanation of it in #10828. Here is an excerpt: The problem here is that the kernel is trying to allocate a contiguous section of 2^7=128 pages, which is 512 KB. This is way too much: kernel pages tend to get fragmented over time and kernel developers often go to great lengths to try allocating at most only 1 contiguous page at a time whenever they can. From the error message, it looks like the culprit is unionfs, but this is misleading: unionfs is the name of the userspace process that was running when the system ran out of memory, but it wasn't unionfs who was allocating the memory: it was the kernel; specifically it was the v9fs_dir_readdir_dotl() function, which is the code for handling the readdir() function in the 9p filesystem (the filesystem that is used to share a directory structure between a qemu host and its VM). If you look at the code, here's what it's doing at the moment it tries to allocate memory: buflen = fid->clnt->msize - P9_IOHDRSZ; rdir = v9fs_alloc_rdir_buf(file, buflen); If you look into v9fs_alloc_rdir_buf(), you will see that it will try to allocate a contiguous buffer of memory (using kzalloc(), which is a wrapper around kmalloc()) of size buflen + 8 bytes or so. So in reality, this code actually allocates a buffer of size proportional to fid->clnt->msize. What is this msize? If you follow the definition of the structures, you will see that it's the negotiated buffer transfer size between 9p client and 9p server. On the client side, it can be controlled with the msize mount option. What this all means is that, the reason for running out of memory is that the code (which we can't easily change) tries to allocate a contiguous buffer of size more or less equal to "negotiated 9p protocol buffer size", which seems to be way too big (in our NixOS tests, at least). After that initial finding, @lethalman tested the gnome3 gdm test without setting the msize parameter at all and it seems to have resolved the problem. The reason why I'm committing this without testing against all of the NixOS VM test is basically that I think we can only go better but not worse than the current state. Signed-off-by: aszlig <aszlig@redmoonstudios.org>
2015-12-14 16:26:24 +00:00
options = "trans=virtio,version=9p2000.L,cache=loose";
neededForBoot = true;
};
"/tmp/xchg" =
{ device = "xchg";
fsType = "9p";
nixos/vm-tests: Remove msize mount option This seems to be the root cause of the random page allocation failures and @wizeman did a very good job on not only finding the root problem but also giving a detailed explanation of it in #10828. Here is an excerpt: The problem here is that the kernel is trying to allocate a contiguous section of 2^7=128 pages, which is 512 KB. This is way too much: kernel pages tend to get fragmented over time and kernel developers often go to great lengths to try allocating at most only 1 contiguous page at a time whenever they can. From the error message, it looks like the culprit is unionfs, but this is misleading: unionfs is the name of the userspace process that was running when the system ran out of memory, but it wasn't unionfs who was allocating the memory: it was the kernel; specifically it was the v9fs_dir_readdir_dotl() function, which is the code for handling the readdir() function in the 9p filesystem (the filesystem that is used to share a directory structure between a qemu host and its VM). If you look at the code, here's what it's doing at the moment it tries to allocate memory: buflen = fid->clnt->msize - P9_IOHDRSZ; rdir = v9fs_alloc_rdir_buf(file, buflen); If you look into v9fs_alloc_rdir_buf(), you will see that it will try to allocate a contiguous buffer of memory (using kzalloc(), which is a wrapper around kmalloc()) of size buflen + 8 bytes or so. So in reality, this code actually allocates a buffer of size proportional to fid->clnt->msize. What is this msize? If you follow the definition of the structures, you will see that it's the negotiated buffer transfer size between 9p client and 9p server. On the client side, it can be controlled with the msize mount option. What this all means is that, the reason for running out of memory is that the code (which we can't easily change) tries to allocate a contiguous buffer of size more or less equal to "negotiated 9p protocol buffer size", which seems to be way too big (in our NixOS tests, at least). After that initial finding, @lethalman tested the gnome3 gdm test without setting the msize parameter at all and it seems to have resolved the problem. The reason why I'm committing this without testing against all of the NixOS VM test is basically that I think we can only go better but not worse than the current state. Signed-off-by: aszlig <aszlig@redmoonstudios.org>
2015-12-14 16:26:24 +00:00
options = "trans=virtio,version=9p2000.L,cache=loose";
neededForBoot = true;
};
"/tmp/shared" =
{ device = "shared";
fsType = "9p";
nixos/vm-tests: Remove msize mount option This seems to be the root cause of the random page allocation failures and @wizeman did a very good job on not only finding the root problem but also giving a detailed explanation of it in #10828. Here is an excerpt: The problem here is that the kernel is trying to allocate a contiguous section of 2^7=128 pages, which is 512 KB. This is way too much: kernel pages tend to get fragmented over time and kernel developers often go to great lengths to try allocating at most only 1 contiguous page at a time whenever they can. From the error message, it looks like the culprit is unionfs, but this is misleading: unionfs is the name of the userspace process that was running when the system ran out of memory, but it wasn't unionfs who was allocating the memory: it was the kernel; specifically it was the v9fs_dir_readdir_dotl() function, which is the code for handling the readdir() function in the 9p filesystem (the filesystem that is used to share a directory structure between a qemu host and its VM). If you look at the code, here's what it's doing at the moment it tries to allocate memory: buflen = fid->clnt->msize - P9_IOHDRSZ; rdir = v9fs_alloc_rdir_buf(file, buflen); If you look into v9fs_alloc_rdir_buf(), you will see that it will try to allocate a contiguous buffer of memory (using kzalloc(), which is a wrapper around kmalloc()) of size buflen + 8 bytes or so. So in reality, this code actually allocates a buffer of size proportional to fid->clnt->msize. What is this msize? If you follow the definition of the structures, you will see that it's the negotiated buffer transfer size between 9p client and 9p server. On the client side, it can be controlled with the msize mount option. What this all means is that, the reason for running out of memory is that the code (which we can't easily change) tries to allocate a contiguous buffer of size more or less equal to "negotiated 9p protocol buffer size", which seems to be way too big (in our NixOS tests, at least). After that initial finding, @lethalman tested the gnome3 gdm test without setting the msize parameter at all and it seems to have resolved the problem. The reason why I'm committing this without testing against all of the NixOS VM test is basically that I think we can only go better but not worse than the current state. Signed-off-by: aszlig <aszlig@redmoonstudios.org>
2015-12-14 16:26:24 +00:00
options = "trans=virtio,version=9p2000.L";
neededForBoot = true;
};
} // optionalAttrs cfg.writableStore
{ "/nix/store" =
{ fsType = "unionfs-fuse";
device = "unionfs";
options = "allow_other,cow,nonempty,chroot=/mnt-root,max_files=32768,hide_meta_files,dirs=/nix/.rw-store=rw:/nix/.ro-store=ro";
};
} // optionalAttrs (cfg.writableStore && cfg.writableStoreUseTmpfs)
{ "/nix/.rw-store" =
{ fsType = "tmpfs";
options = "mode=0755";
neededForBoot = true;
};
} // optionalAttrs cfg.useBootLoader
{ "/boot" =
{ device = "/dev/vdb2";
fsType = "vfat";
options = "ro";
noCheck = true; # fsck fails on a r/o filesystem
};
});
swapDevices = mkVMOverride [ ];
boot.initrd.luks.devices = mkVMOverride [];
# Don't run ntpd in the guest. It should get the correct time from KVM.
services.ntp.enable = false;
system.build.vm = pkgs.runCommand "nixos-vm" { preferLocalBuild = true; }
''
mkdir -p $out/bin
ln -s ${config.system.build.toplevel} $out/system
ln -s ${pkgs.writeScript "run-nixos-vm" startVM} $out/bin/run-${vmName}-vm
'';
# When building a regular system configuration, override whatever
# video driver the host uses.
services.xserver.videoDrivers = mkVMOverride [ "modesetting" ];
services.xserver.defaultDepth = mkVMOverride 0;
services.xserver.resolutions = mkVMOverride [ { x = 1024; y = 768; } ];
services.xserver.monitorSection =
''
# Set a higher refresh rate so that resolutions > 800x600 work.
HorizSync 30-140
VertRefresh 50-160
'';
# Wireless won't work in the VM.
networking.wireless.enable = mkVMOverride false;
networking.connman.enable = mkVMOverride false;
# Speed up booting by not waiting for ARP.
networking.dhcpcd.extraConfig = "noarp";
networking.usePredictableInterfaceNames = false;
system.requiredKernelConfig = with config.lib.kernelConfig;
[ (isEnabled "VIRTIO_BLK")
(isEnabled "VIRTIO_PCI")
(isEnabled "VIRTIO_NET")
(isEnabled "EXT4_FS")
(isYes "BLK_DEV")
(isYes "PCI")
(isYes "EXPERIMENTAL")
(isYes "NETDEVICES")
(isYes "NET_CORE")
(isYes "INET")
(isYes "NETWORK_FILESYSTEMS")
] ++ optional (!cfg.graphics) [
(isYes "SERIAL_8250_CONSOLE")
(isYes "SERIAL_8250")
];
};
}