build-vm-kvm

#
# kvm/qemu specific functions
#
################################################################
#
# Copyright (c) 1995-2014 SUSE Linux Products GmbH
# Copyright (c) 2021 SUSE LLC
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License version 2 or 3 as
# published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program (see the file COPYING); if not, write to the
# Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
#
################################################################

kvm_bin=/usr/bin/qemu-kvm
test ! -x $kvm_bin  -a -x /usr/bin/kvm && kvm_bin=/usr/bin/kvm
kvm_console=ttyS0

# assume virtio support by default
kvm_device=virtio-blk-pci
kvm_device_opts=
kvm_drive_opts="cache=unsafe"
kvm_serial_device=
kvm_rng_device=virtio-rng-pci
kvm_options=
kvm_qemu_append="no-kvmclock"
kvm_cpu="-cpu host"

kvm_check_ppc970() {
    if ! grep -q -E '(kvm_rma_count.*kvm_hpt_count)|(kvm_hpt_count.*kvm_rma_count)' /proc/cmdline ; then 
	cleanup_and_exit 3 "put kvm_rma_count=<VM number> or kvm_hpt_count=<> to your boot options"
    fi
}

kvm_check_hugetlb() {
    if ! grep -q "$HUGETLBFSPATH" /proc/mounts ; then
	cleanup_and_exit 4 "hugetlbfs is not mounted to $HUGETLBFSPATH"
    fi
    local HUGETLBBLKSIZE=$(stat -f -c "%S" "$HUGETLBFSPATH")
    HUGETLBBLKSIZE=$(( ${HUGETLBBLKSIZE:-0} / 1024 ))
    if test "$HUGETLBBLKSIZE" -lt 1 -o ! -e  "/sys/kernel/mm/hugepages/hugepages-${HUGETLBBLKSIZE}kB" ; then
	cleanup_and_exit 4 "could not determine hugetlbfs block size"
    fi
    local PAGES_FREE=$(cat /sys/kernel/mm/hugepages/hugepages-${HUGETLBBLKSIZE}kB/free_hugepages)
    local PAGES_REQ=$(( ${VM_MEMSIZE:-64} * 1024 / $HUGETLBBLKSIZE ))
    if test "$PAGES_FREE" -lt "$PAGES_REQ" ; then
	echo "expected $PAGES_REQ to be available (have $PAGES_FREE)"
	echo "please adjust nr_hugepages"
	cleanup_and_exit 4
    fi
}

vm_verify_options_kvm() {
    if test -n "$KILL" -o -n "$DO_WIPE" ; then
	return
    fi  

    vm_kernel=
    vm_initrd=
    vm_cmdline=
    
    # newer Ubuntu versions have only kvm executable
    if test -x "/usr/bin/kvm" -a ! -x "$kvm_bin"; then
	kvm_bin="/usr/bin/kvm"
    fi

    # overwrite some options for specific host architectures
    case `uname -m` in
	armv7l)
	    kvm_bin="/usr/bin/qemu-system-arm"
	    kvm_console=ttyAMA0
	    kvm_options="-enable-kvm -M virt"
	    vm_kernel=/boot/zImage
	    vm_initrd=/boot/initrd
	    test -e /boot/kernel.obs.guest && vm_kernel=/boot/kernel.obs.guest
	    test -e /boot/initrd.obs.guest && vm_initrd=/boot/initrd.obs.guest
	    # prefer the guest kernel/initrd
	    test -e /boot/zImage.guest && vm_kernel=/boot/zImage.guest
	    test -e /boot/initrd.guest && vm_initrd=/boot/initrd.guest
	    kvm_device=virtio-blk-device
	    kvm_rng_device=virtio-rng-device
	    ;;
	armv8l|aarch64)
	    kvm_bin="/usr/bin/qemu-system-aarch64"
	    kvm_console=ttyAMA0
	    kvm_options="-enable-kvm"
	    vm_kernel=/boot/Image
	    vm_initrd=/boot/initrd
	    test -e /boot/kernel.obs.guest && vm_kernel=/boot/kernel.obs.guest
	    test -e /boot/initrd.obs.guest && vm_initrd=/boot/initrd.obs.guest
	    if test "${BUILD_ARCH#aarch}" != "$BUILD_ARCH" -o "${BUILD_ARCH#armv8}" != "$BUILD_ARCH"; then
		test -e /boot/Image.guest && vm_kernel=/boot/Image.guest
		test -e /boot/initrd.guest && vm_initrd=/boot/initrd.guest
	    else
		# Running an armv7 kernel on aarch64
		# prefer the guest kernel/initrd
		test -e /boot/Image.guest32 && vm_kernel=/boot/Image.guest32
		test -e /boot/initrd.guest32 && vm_initrd=/boot/initrd.guest32
	    fi
	    # We want to use the host gic version in order to make use
	    # of all available features (e.g. more than 8 CPUs) and avoid
	    # the emulation overhead of vGICv2 on a GICv3 host.
            # move IO into separate I/O thread
	    kvm_options="$kvm_options -M virt,accel=kvm,usb=off,gic-version=host -sandbox on -object iothread,id=io0"
	    kvm_device=virtio-blk-device
	    kvm_device_opts=",iothread=io0"
	    kvm_rng_device=virtio-rng-device
	    ;;
	ppc|ppcle|ppc64|ppc64le)
	    kvm_bin="/usr/bin/qemu-system-ppc64"
	    kvm_console=hvc0
	    kvm_options="-enable-kvm -M pseries"
	    if grep -q POWER8 /proc/cpuinfo ; then
	        # This option only exists with QEMU 5.0 or newer
		if $kvm_bin -machine 'pseries,?' 2>&1 | grep -q cap-ccf-assist ; then
		kvm_options="-enable-kvm -M pseries,cap-ccf-assist=off"
		fi
	    fi
	    if grep -q "POWER9" /proc/cpuinfo; then
		# This option has been added in QEMU 2.10
		if $kvm_bin -machine 'help' 2>&1 | grep -q ^pseries-6; then
		    kvm_options="-enable-kvm -M pseries,max-cpu-compat=power8"
		else
		    kvm_cpu="-cpu host,compat=power8"
		fi
	    fi
	    grep -q PPC970MP /proc/cpuinfo && kvm_check_ppc970
	    vm_kernel=/boot/vmlinux
	    vm_initrd=/boot/initrd
	    test -e /boot/kernel.obs.guest && vm_kernel=/boot/kernel.obs.guest
	    test -e /boot/initrd.obs.guest && vm_initrd=/boot/initrd.obs.guest
	    if test "$BUILD_ARCH" = ppc64le -a -e /boot/vmlinuxle ; then
		vm_kernel=/boot/vmlinuxle
		vm_initrd=/boot/initrdle
	    fi
	    if test -e /boot/vmlinuxbe -a -e /boot/initrdbe ; then
		if test "$BUILD_ARCH" = ppc -o "$BUILD_ARCH" = ppc64 ; then
		    vm_kernel=/boot/vmlinuxbe
		    vm_initrd=/boot/initrdbe
		fi
	    fi
	    grep -q "pSeries" /proc/cpuinfo && kvm_device=scsi-hd	# no virtio on pSeries
	    grep -q "PowerNV" /proc/cpuinfo || kvm_device=scsi-hd	# no virtio on ppc != power7 yet
	    ;;
	s390|s390x)
	    kvm_bin="/usr/bin/qemu-system-s390x"
	    kvm_options="-enable-kvm"
	    kvm_console=hvc0
	    vm_kernel=/boot/image
	    vm_initrd=/boot/initrd
	    test -e /boot/kernel.obs.guest && vm_kernel=/boot/kernel.obs.guest
	    test -e /boot/initrd.obs.guest && vm_initrd=/boot/initrd.obs.guest
	    kvm_device=virtio-blk-ccw
	    kvm_serial_device=virtio-serial-ccw
	    kvm_rng_device=virtio-rng-ccw
	    ;;
	i586|i686|x86_64)
	    test ! -x $kvm_bin -a -x /usr/bin/qemu-system-x86_64 && kvm_bin=/usr/bin/qemu-system-x86_64

            # build minimal machine - TODO enable sandbox on all architectures
            # options to evaluate: mem-merge=off
            kvm_options="-M pc,accel=kvm,usb=off,dump-guest-core=off,vmport=off -sandbox on"

            # from qemu-microvm, minified and hardened PC bios (slightly faster than seabios)
            test -e /usr/share/qemu/qboot.rom && kvm_options="$kvm_options -bios /usr/share/qemu/qboot.rom"

            # move IO into separate I/O thread
	    if test "${kvm_device%%-*}" = "virtio" ; then
                    kvm_options="$kvm_options -object iothread,id=io0"
		    kvm_device_opts=",iothread=io0"
            fi

            # TODO appears to be unstable
	    # $kvm_bin --help | grep -q io_uring && kvm_drive_opts="$kvm_drive_opts,aio=io_uring"
            ;;
    esac

    # check if we can run kvm
    if ! test -r /dev/kvm ; then
	cleanup_and_exit 4 "kvm device not accessible: either the kvm kernel-module is not loaded or hardware virtualization is deactivated in the BIOS."
    fi
    if ! test -x "$kvm_bin" ; then
	cleanup_and_exit 4 "$kvm_bin is not installed"
    fi

    # check hugepages
    test -n "$HUGETLBFSPATH" -a "$VM_TYPE" = kvm && kvm_check_hugetlb

    # set kernel
    test -n "$VM_KERNEL" && vm_kernel="$VM_KERNEL"
    if test -z "$vm_kernel" ; then
	vm_kernel=/boot/vmlinuz
	test -e /boot/kernel.obs.guest && vm_kernel=/boot/kernel.obs.guest
    fi

    # set initrd
    test -n "$VM_INITRD" && vm_initrd="$VM_INITRD"
    test -n "$VM_CMDLINE" && vm_cmdline="$VM_CMDLINE"
    if test -z "$vm_initrd" ; then
	# find a nice default
	if test -e /boot/initrd.obs.guest ; then
	    vm_initrd=/boot/initrd.obs.guest
	elif test -e "/boot/initrd-build" ; then
	    vm_initrd="/boot/initrd-build"
	elif test -e "/boot/initrd-virtio" ; then
	    vm_initrd="/boot/initrd-virtio"
	else
	    vm_initrd="/boot/initrd"
	    kvm_device=ide-hd
	    # use /etc/sysconfig/kernel as indication if we have virtio
	    if test -e /etc/sysconfig/kernel ; then
		local im=$(INITRD_MODULES=; . /etc/sysconfig/kernel; echo "$INITRD_MODULES")
		if test "$im" != "${im/virtio/}" ; then
		    kvm_device=virtio-blk-pci
		fi
	    fi
	fi
    fi

    case $kvm_device in
	virtio*)
	    VM_ROOTDEV=/dev/disk/by-id/virtio-0
	    VM_SWAPDEV=/dev/disk/by-id/virtio-1
	    kvm_qemu_append="kvmclock"
	    ;;
	*)
	    VM_ROOTDEV=/dev/sda
	    VM_SWAPDEV=/dev/sdb
	    ;;
    esac

    if test -n "$VM_NETOPT" -o -n "$VM_NETDEVOPT" ; then
        if test -n "$VM_NETOPT" ; then
           for item in "${VM_NETOPT[@]}" ; do
              kvm_options="$kvm_options -net $item"
           done
        fi
        if test -n "$VM_NETDEVOPT" ; then
           for item in "${VM_NETDEVOPT[@]}" ; do
              kvm_options="$kvm_options -netdev $item"
           done
        fi
    fi

    if test -n "$VM_DEVICEOPT" ; then
        for item in "${VM_DEVICEOPT[@]}" ; do
            kvm_options="$kvm_options -device $item"
        done
    fi

    if test -n "$kvm_rng_device" ; then
        if test -c /dev/hwrng &&
           test -w /dev/hwrng &&
           test -f /sys/class/misc/hw_random/rng_current &&
           test "$(cat /sys/class/misc/hw_random/rng_current)" != none; then
            rng_dev="/dev/hwrng"
        else
            rng_dev="/dev/random"
        fi
        kvm_options="$kvm_options -object rng-random,filename=$rng_dev,id=rng0 -device $kvm_rng_device,rng=rng0"
    fi
}

kvm_add_console_args() {
   local serial_device="$1"
    # the serial console device needs to be compiled into the target kernel
    # which is why we can not use virtio-serial on other platforms
    if test -n "$serial_device" ; then
	if test -n "$VM_CONSOLE_INPUT" ; then
	    echo "Using virtio-serial support and enabling console input"
	    qemu_args=("${qemu_args[@]}" -device "$serial_device" -device virtconsole,chardev=virtiocon0 -chardev stdio,mux=on,id=virtiocon0 -mon chardev=virtiocon0)
	else
	    echo "Using virtio-serial support"
	    qemu_args=("${qemu_args[@]}" -device "$serial_device" -device virtconsole,chardev=virtiocon0 -chardev stdio,id=virtiocon0)
	fi
    elif test -n "$VM_CONSOLE_INPUT" ; then
	echo "Enabling console input"
	qemu_args=("${qemu_args[@]}" -serial mon:stdio)
    else
	echo "Using UART console"
	qemu_args=("${qemu_args[@]}" -serial stdio)
    fi

    # setup the monitor
    if ! test -e "${VM_ROOT}.qemu/monitor"; then
      mkdir -p "${VM_ROOT}.qemu"
      mkfifo "${VM_ROOT}.qemu/monitor"
      test -n "$VM_USER" && chown "$VM_USER" "${VM_ROOT}.qemu"
    fi
    qemu_args=("${qemu_args[@]}" -chardev socket,id=monitor,server=on,wait=off,path="${VM_ROOT}.qemu/monitor" -mon chardev=monitor,mode=readline)
}

vm_startup_kvm() {
    qemu_bin="$kvm_bin"
    qemu_args=(-drive file="$VM_ROOT",format=raw,if=none,id=disk,$kvm_drive_opts -device "$kvm_device$kvm_device_opts",drive=disk,serial=0)
    if [ -n "$VM_USER" ] ; then
	getent passwd "$VM_USER" > /dev/null || cleanup_and_exit 3 "cannot find KVM user '$VM_USER'"
    elif test $UID = 0 ; then
	# use qemu user by default if available
	getent passwd qemu >/dev/null && VM_USER=qemu
    fi
    [ -n "$VM_USER" ] && kvm_options="$kvm_options -runas $VM_USER"
    if test -n "$VM_SWAP" ; then
	qemu_args=("${qemu_args[@]}" -drive file="$VM_SWAP",format=raw,if=none,id=swap,$kvm_drive_opts -device "$kvm_device$kvm_device_opts",drive=swap,serial=1)
    fi

    kvm_add_console_args "$kvm_serial_device"

    # figure out kvm cpu based on possible provided hostarch file
    case `uname -m` in
        armv8l|aarch64)
            if test "$BUILD_HOST_ARCH" != "aarch64" ; then
                kvm_cpu="-cpu host,aarch64=off"
            fi
            ;;
    esac

    if test -n "$BUILD_JOBS" -a "$icecream" = 0 -a -z "$BUILD_THREADS" ; then
	qemu_args=("${qemu_args[@]}" "-smp" "$BUILD_JOBS")
    elif test -n "$BUILD_JOBS" -a -n "$BUILD_THREADS" ; then
	qemu_args=("${qemu_args[@]}" "-smp" "$BUILD_JOBS,threads=$BUILD_THREADS")
    fi
    if test "$VM_TYPE" = kvm ; then
	test -n "$HUGETLBFSPATH" && kvm_options="$kvm_options -mem-prealloc -mem-path $HUGETLBFSPATH"
    fi
    qemu_append="root=$VM_ROOTDEV"
    if test -n "$VMDISK_FILESYSTEM" ; then
        qemu_append="$qemu_append rootfstype=$VMDISK_FILESYSTEM"
    fi
    if test -n "$VMDISK_MOUNT_OPTIONS" ; then
        qemu_append="$qemu_append rootflags=${VMDISK_MOUNT_OPTIONS#-o }"
    fi
    if test -n "$vm_cmdline" ; then
        qemu_append="$qemu_append $vm_cmdline"
    else
        # Pick sensible defaults
        qemu_append="$qemu_append $kvm_qemu_append $vm_linux_kernel_parameter elevator=noop"
        qemu_append="$qemu_append nmi_watchdog=0 rw rd.driver.pre=binfmt_misc"
    fi
    qemu_append="$qemu_append $vm_linux_always_append console=$kvm_console init=$vm_init_script"
    if test -z "$VM_NETOPT" -a -z "$VM_NETDEVOPT"; then
        kvm_options="$kvm_options -net none"
    fi
    if test -n "$VM_TELNET"; then
        kvm_options="$kvm_options -netdev user,id=telnet,hostfwd=tcp:127.0.0.1:$VM_TELNET-:23 -device e1000,netdev=telnet"
    fi
    if test -n "$VM_CUSTOMOPT"; then
        kvm_options="$kvm_options $VM_CUSTOMOPT"
    fi
    set -- $qemu_bin -nodefaults -no-reboot -nographic -vga none $kvm_cpu $kvm_options \
	-kernel $vm_kernel \
	-initrd $vm_initrd \
	-append "$qemu_append" \
	${VM_MEMSIZE:+-m $VM_MEMSIZE} \
	"${qemu_args[@]}"

    if test "$PERSONALITY" != 0 ; then
	# have to switch back to PER_LINUX to make qemu work
	set -- linux64 "$@"
    fi
    export QEMU_AUDIO_DRV=none		# we do not want to have sound inside the VMs
    echo "$@"
    "$@"
    qemu_ret=$?
    test "$qemu_ret" = "137" && cleanup_and_exit 3 "qemu got SIGKILL"
    test "$qemu_ret" = "143" && cleanup_and_exit 1 "qemu got SIGTERM"
}

vm_kill_kvm() {
    if ! fuser -k -TERM "$VM_ROOT" ; then
	echo "could not kill build in $VM_ROOT, kvm instance is dead already"
    fi
}

vm_fixup_kvm() {
    # check if we will use a kernel from the build root, in this case
    # we assume the kernel does virtio
    if test -z "$VM_KERNEL" -a -e "$BUILD_ROOT/.build.kernel.$VM_TYPE" ; then
	# ide-hd is the non-virtio default
	if test "$kvm_device" = ide-hd ; then
	    kvm_device=virtio-blk-pci
	    VM_ROOTDEV=/dev/disk/by-id/virtio-0
	    VM_SWAPDEV=/dev/disk/by-id/virtio-1
	    kvm_qemu_append="kvmclock"
	fi
    fi
    if test "$VM_TYPE" = kvm -a -z "$kvm_serial_device" ; then
	if test -f "$BUILD_ROOT/.build.console.kvm" -a ! -L "$BUILD_ROOT/.build.console.kvm" && grep -q '^virtio$' "$BUILD_ROOT/.build.console.kvm" ; then
	    echo "Detected virtio-serial support"
	    kvm_serial_device=virtio-serial,max_ports=2
	    kvm_console=hvc0
	fi
    fi
}

vm_attach_root_kvm() {
    :
}

vm_attach_swap_kvm() {
    :
}

vm_detach_root_kvm() {
    :
}

vm_detach_swap_kvm() {
    :
}

vm_cleanup_kvm() {
    :
}

vm_sysrq_kvm() {
    perl -e 'use Socket; socket(SOCK, PF_UNIX, SOCK_STREAM, 0) || die("socket: $!\n"); 
             connect(SOCK, sockaddr_un($ARGV[0])) || die("connect: $!\n");
             syswrite(SOCK, "$ARGV[1]\n");
             shutdown(SOCK, 1);
             my $output;
             read(SOCK, $output, 8192);' "$VM_ROOT.qemu/monitor" "sendkey alt-print-$1"
}

vm_wipe_kvm() {
    :
}