1 package PVE
::QemuServer
;
11 use File
::Copy
qw(copy);
22 use List
::Util
qw(first);
25 use Storable
qw(dclone);
26 use Time
::HiRes
qw(gettimeofday usleep);
30 use PVE
::Cluster
qw(cfs_register_file cfs_read_file cfs_write_file);
33 use PVE
::DataCenterConfig
;
34 use PVE
::Exception
qw(raise raise_param_exc);
35 use PVE
::Format
qw(render_duration render_bytes);
36 use PVE
::GuestHelpers
qw(safe_string_ne safe_num_ne safe_boolean_ne);
37 use PVE
::Mapping
::PCI
;
38 use PVE
::Mapping
::USB
;
40 use PVE
::JSONSchema
qw(get_standard_option parse_property_string);
43 use PVE
::RESTEnvironment
qw(log_warn);
44 use PVE
::RPCEnvironment
;
48 use PVE
::Tools
qw(run_command file_read_firstline file_get_contents dir_glob_foreach get_host_arch $IPV6RE);
52 use PVE
::QemuServer
::Helpers
qw(min_version config_aware_timeout windows_version);
53 use PVE
::QemuServer
::Cloudinit
;
54 use PVE
::QemuServer
::CGroup
;
55 use PVE
::QemuServer
::CPUConfig
qw(print_cpu_device get_cpu_options);
56 use PVE
::QemuServer
::Drive
qw(is_valid_drivename drive_is_cloudinit drive_is_cdrom drive_is_read_only parse_drive print_drive);
57 use PVE
::QemuServer
::Machine
;
58 use PVE
::QemuServer
::Memory
;
59 use PVE
::QemuServer
::Monitor
qw(mon_cmd);
60 use PVE
::QemuServer
::PCI
qw(print_pci_addr print_pcie_addr print_pcie_root_port parse_hostpci);
61 use PVE
::QemuServer
::USB
;
65 require PVE
::Network
::SDN
::Zones
;
69 my $EDK2_FW_BASE = '/usr/share/pve-edk2-firmware/';
73 "$EDK2_FW_BASE/OVMF_CODE_4M.fd",
74 "$EDK2_FW_BASE/OVMF_VARS_4M.fd",
77 "$EDK2_FW_BASE/OVMF_CODE_4M.fd",
78 "$EDK2_FW_BASE/OVMF_VARS_4M.ms.fd",
81 "$EDK2_FW_BASE/OVMF_CODE_4M.secboot.fd",
82 "$EDK2_FW_BASE/OVMF_VARS_4M.fd",
85 "$EDK2_FW_BASE/OVMF_CODE_4M.secboot.fd",
86 "$EDK2_FW_BASE/OVMF_VARS_4M.ms.fd",
89 "$EDK2_FW_BASE/OVMF_CODE.fd",
90 "$EDK2_FW_BASE/OVMF_VARS.fd",
95 "$EDK2_FW_BASE/AAVMF_CODE.fd",
96 "$EDK2_FW_BASE/AAVMF_VARS.fd",
101 my $cpuinfo = PVE
::ProcFSTools
::read_cpuinfo
();
103 # Note about locking: we use flock on the config file protect against concurent actions.
104 # Aditionaly, we have a 'lock' setting in the config file. This can be set to 'migrate',
105 # 'backup', 'snapshot' or 'rollback'. Most actions are not allowed when such lock is set.
106 # But you can ignore this kind of lock with the --skiplock flag.
114 PVE
::JSONSchema
::register_standard_option
('pve-qm-stateuri', {
115 description
=> "Some command save/restore state from this location.",
121 PVE
::JSONSchema
::register_standard_option
('pve-qemu-machine', {
122 description
=> "Specifies the QEMU machine type.",
124 pattern
=> '(pc|pc(-i440fx)?-\d+(\.\d+)+(\+pve\d+)?(\.pxe)?|q35|pc-q35-\d+(\.\d+)+(\+pve\d+)?(\.pxe)?|virt(?:-\d+(\.\d+)+)?(\+pve\d+)?)',
129 # FIXME: remove in favor of just using the INotify one, it's cached there exactly the same way
132 $nodename_cache //= PVE
::INotify
::nodename
();
133 return $nodename_cache;
140 enum
=> [qw(i6300esb ib700)],
141 description
=> "Watchdog type to emulate.",
142 default => 'i6300esb',
147 enum
=> [qw(reset shutdown poweroff pause debug none)],
148 description
=> "The action to perform if after activation the guest fails to poll the watchdog in time.",
152 PVE
::JSONSchema
::register_format
('pve-qm-watchdog', $watchdog_fmt);
156 description
=> "Enable/disable communication with a QEMU Guest Agent (QGA) running in the VM.",
161 fstrim_cloned_disks
=> {
162 description
=> "Run fstrim after moving a disk or migrating the VM.",
167 'freeze-fs-on-backup' => {
168 description
=> "Freeze/thaw guest filesystems on backup for consistency.",
174 description
=> "Select the agent type",
178 enum
=> [qw(virtio isa)],
184 description
=> "Select the VGA type.",
189 enum
=> [qw(cirrus qxl qxl2 qxl3 qxl4 none serial0 serial1 serial2 serial3 std virtio virtio-gl vmware)],
192 description
=> "Sets the VGA memory (in MiB). Has no effect with serial display.",
204 description
=> "The size of the file in MB.",
208 pattern
=> '[a-zA-Z0-9\-]+',
210 format_description
=> 'string',
211 description
=> "The name of the file. Will be prefixed with 'pve-shm-'. Default is the VMID. Will be deleted when the VM is stopped.",
218 enum
=> [qw(ich9-intel-hda intel-hda AC97)],
219 description
=> "Configure an audio device."
223 enum
=> ['spice', 'none'],
226 description
=> "Driver backend for the audio device."
230 my $spice_enhancements_fmt = {
235 description
=> "Enable folder sharing via SPICE. Needs Spice-WebDAV daemon installed in the VM."
239 enum
=> ['off', 'all', 'filter'],
242 description
=> "Enable video streaming. Uses compression for detected video streams."
249 enum
=> ['/dev/urandom', '/dev/random', '/dev/hwrng'],
251 description
=> "The file on the host to gather entropy from. In most cases '/dev/urandom'"
252 ." should be preferred over '/dev/random' to avoid entropy-starvation issues on the"
253 ." host. Using urandom does *not* decrease security in any meaningful way, as it's"
254 ." still seeded from real entropy, and the bytes provided will most likely be mixed"
255 ." with real entropy on the guest as well. '/dev/hwrng' can be used to pass through"
256 ." a hardware RNG from the host.",
260 description
=> "Maximum bytes of entropy allowed to get injected into the guest every"
261 ." 'period' milliseconds. Prefer a lower value when using '/dev/random' as source. Use"
262 ." `0` to disable limiting (potentially dangerous!).",
265 # default is 1 KiB/s, provides enough entropy to the guest to avoid boot-starvation issues
266 # (e.g. systemd etc...) while allowing no chance of overwhelming the host, provided we're
267 # reading from /dev/urandom
272 description
=> "Every 'period' milliseconds the entropy-injection quota is reset, allowing"
273 ." the guest to retrieve another 'max_bytes' of entropy.",
279 my $meta_info_fmt = {
282 description
=> "The guest creation timestamp as UNIX epoch time",
288 description
=> "The QEMU (machine) version from the time this VM was created.",
289 pattern
=> '\d+(\.\d+)+',
298 description
=> "Specifies whether a VM will be started during system bootup.",
304 description
=> "Automatic restart after crash (currently ignored).",
309 type
=> 'string', format
=> 'pve-hotplug-features',
310 description
=> "Selectively enable hotplug features. This is a comma separated list of"
311 ." hotplug features: 'network', 'disk', 'cpu', 'memory', 'usb' and 'cloudinit'. Use '0' to disable"
312 ." hotplug completely. Using '1' as value is an alias for the default `network,disk,usb`."
313 ." USB hotplugging is possible for guests with machine version >= 7.1 and ostype l26 or"
315 default => 'network,disk,usb',
320 description
=> "Allow reboot. If set to '0' the VM exit on reboot.",
326 description
=> "Lock/unlock the VM.",
327 enum
=> [qw(backup clone create migrate rollback snapshot snapshot-delete suspending suspended)],
332 description
=> "Limit of CPU usage.",
333 verbose_description
=> "Limit of CPU usage.\n\nNOTE: If the computer has 2 CPUs, it has"
334 ." total of '2' CPU time. Value '0' indicates no CPU limit.",
342 description
=> "CPU weight for a VM, will be clamped to [1, 10000] in cgroup v2.",
343 verbose_description
=> "CPU weight for a VM. Argument is used in the kernel fair scheduler."
344 ." The larger the number is, the more CPU time this VM gets. Number is relative to"
345 ." weights of all the other running VMs.",
348 default => 'cgroup v1: 1024, cgroup v2: 100',
353 description
=> "Amount of RAM for the VM in MiB. This is the maximum available memory when"
354 ." you use the balloon device.",
361 description
=> "Amount of target RAM for the VM in MiB. Using zero disables the ballon driver.",
367 description
=> "Amount of memory shares for auto-ballooning. The larger the number is, the"
368 ." more memory this VM gets. Number is relative to weights of all other running VMs."
369 ." Using zero disables auto-ballooning. Auto-ballooning is done by pvestatd.",
377 description
=> "Keyboard layout for VNC server. This option is generally not required and"
378 ." is often better handled from within the guest OS.",
379 enum
=> PVE
::Tools
::kvmkeymaplist
(),
384 type
=> 'string', format
=> 'dns-name',
385 description
=> "Set a name for the VM. Only used on the configuration web interface.",
390 description
=> "SCSI controller model",
391 enum
=> [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
397 description
=> "Description for the VM. Shown in the web-interface VM's summary."
398 ." This is saved as comment inside the configuration file.",
399 maxLength
=> 1024 * 8,
404 enum
=> [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 win11 l24 l26 solaris)],
405 description
=> "Specify guest operating system.",
406 verbose_description
=> <<EODESC,
407 Specify guest operating system. This is used to enable special
408 optimization/features for specific operating systems:
411 other;; unspecified OS
412 wxp;; Microsoft Windows XP
413 w2k;; Microsoft Windows 2000
414 w2k3;; Microsoft Windows 2003
415 w2k8;; Microsoft Windows 2008
416 wvista;; Microsoft Windows Vista
417 win7;; Microsoft Windows 7
418 win8;; Microsoft Windows 8/2012/2012r2
419 win10;; Microsoft Windows 10/2016/2019
420 win11;; Microsoft Windows 11/2022
421 l24;; Linux 2.4 Kernel
422 l26;; Linux 2.6 - 6.X Kernel
423 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
428 type
=> 'string', format
=> 'pve-qm-boot',
429 description
=> "Specify guest boot order. Use the 'order=' sub-property as usage with no"
430 ." key or 'legacy=' is deprecated.",
434 type
=> 'string', format
=> 'pve-qm-bootdisk',
435 description
=> "Enable booting from specified disk. Deprecated: Use 'boot: order=foo;bar' instead.",
436 pattern
=> '(ide|sata|scsi|virtio)\d+',
441 description
=> "The number of CPUs. Please use option -sockets instead.",
448 description
=> "The number of CPU sockets.",
455 description
=> "The number of cores per socket.",
462 description
=> "Enable/disable NUMA.",
468 description
=> "Enable/disable hugepages memory.",
469 enum
=> [qw(any 2 1024)],
475 description
=> "Use together with hugepages. If enabled, hugepages will not not be deleted"
476 ." after VM shutdown and can be used for subsequent starts.",
481 description
=> "Number of hotplugged vcpus.",
488 description
=> "Enable/disable ACPI.",
493 description
=> "Enable/disable communication with the QEMU Guest Agent and its properties.",
495 format
=> $agent_fmt,
500 description
=> "Enable/disable KVM hardware virtualization.",
506 description
=> "Enable/disable time drift fix.",
512 description
=> "Set the real time clock (RTC) to local time. This is enabled by default if"
513 ." the `ostype` indicates a Microsoft Windows OS.",
518 description
=> "Freeze CPU at startup (use 'c' monitor command to start execution).",
522 type
=> 'string', format
=> $vga_fmt,
523 description
=> "Configure the VGA hardware.",
524 verbose_description
=> "Configure the VGA Hardware. If you want to use high resolution"
525 ." modes (>= 1280x1024x16) you may need to increase the vga memory option. Since QEMU"
526 ." 2.9 the default VGA display type is 'std' for all OS types besides some Windows"
527 ." versions (XP and older) which use 'cirrus'. The 'qxl' option enables the SPICE"
528 ." display server. For win* OS you can select how many independent displays you want,"
529 ." Linux guests can add displays them self.\nYou can also run without any graphic card,"
530 ." using a serial device as terminal.",
534 type
=> 'string', format
=> 'pve-qm-watchdog',
535 description
=> "Create a virtual hardware watchdog device.",
536 verbose_description
=> "Create a virtual hardware watchdog device. Once enabled (by a guest"
537 ." action), the watchdog must be periodically polled by an agent inside the guest or"
538 ." else the watchdog will reset the guest (or execute the respective action specified)",
543 typetext
=> "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
544 description
=> "Set the initial date of the real time clock. Valid format for date are:"
545 ."'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
546 pattern
=> '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
549 startup
=> get_standard_option
('pve-startup-order'),
553 description
=> "Enable/disable Template.",
559 description
=> "Arbitrary arguments passed to kvm.",
560 verbose_description
=> <<EODESCR,
561 Arbitrary arguments passed to kvm, for example:
563 args: -no-reboot -smbios 'type=0,vendor=FOO'
565 NOTE: this option is for experts only.
572 description
=> "Enable/disable the USB tablet device.",
573 verbose_description
=> "Enable/disable the USB tablet device. This device is usually needed"
574 ." to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with"
575 ." normal VNC clients. If you're running lots of console-only guests on one host, you"
576 ." may consider disabling this to save some context switches. This is turned off by"
577 ." default if you use spice (`qm set <vmid> --vga qxl`).",
582 description
=> "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
586 migrate_downtime
=> {
589 description
=> "Set maximum tolerated downtime (in seconds) for migrations.",
595 type
=> 'string', format
=> 'pve-qm-ide',
596 typetext
=> '<volume>',
597 description
=> "This is an alias for option -ide2",
601 description
=> "Emulated CPU type.",
603 format
=> 'pve-vm-cpu-conf',
605 parent
=> get_standard_option
('pve-snapshot-name', {
607 description
=> "Parent snapshot name. This is used internally, and should not be modified.",
611 description
=> "Timestamp for snapshots.",
617 type
=> 'string', format
=> 'pve-volume-id',
618 description
=> "Reference to a volume which stores the VM state. This is used internally"
621 vmstatestorage
=> get_standard_option
('pve-storage-id', {
622 description
=> "Default storage for VM state volumes/files.",
625 runningmachine
=> get_standard_option
('pve-qemu-machine', {
626 description
=> "Specifies the QEMU machine type of the running vm. This is used internally"
630 description
=> "Specifies the QEMU '-cpu' parameter of the running vm. This is used"
631 ." internally for snapshots.",
634 pattern
=> $PVE::QemuServer
::CPUConfig
::qemu_cmdline_cpu_re
,
635 format_description
=> 'QEMU -cpu parameter'
637 machine
=> get_standard_option
('pve-qemu-machine'),
639 description
=> "Virtual processor architecture. Defaults to the host.",
642 enum
=> [qw(x86_64 aarch64)],
645 description
=> "Specify SMBIOS type 1 fields.",
646 type
=> 'string', format
=> 'pve-qm-smbios1',
653 description
=> "Sets the protection flag of the VM. This will disable the remove VM and"
654 ." remove disk operations.",
660 enum
=> [ qw(seabios ovmf) ],
661 description
=> "Select BIOS implementation.",
662 default => 'seabios',
666 pattern
=> '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
667 format_description
=> 'UUID',
668 description
=> "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0'"
669 ." to disable explicitly.",
670 verbose_description
=> "The VM generation ID (vmgenid) device exposes a 128-bit integer"
671 ." value identifier to the guest OS. This allows to notify the guest operating system"
672 ." when the virtual machine is executed with a different configuration (e.g. snapshot"
673 ." execution or creation from a template). The guest operating system notices the"
674 ." change, and is then able to react as appropriate by marking its copies of"
675 ." distributed databases as dirty, re-initializing its random number generator, etc.\n"
676 ."Note that auto-creation only works when done through API/CLI create or update methods"
677 .", but not when manually editing the config file.",
678 default => "1 (autogenerated)",
683 format
=> 'pve-volume-id',
685 description
=> "Script that will be executed during various steps in the vms lifetime.",
689 format
=> $ivshmem_fmt,
690 description
=> "Inter-VM shared memory. Useful for direct communication between VMs, or to"
696 format
=> $audio_fmt,
697 description
=> "Configure a audio device, useful in combination with QXL/Spice.",
700 spice_enhancements
=> {
702 format
=> $spice_enhancements_fmt,
703 description
=> "Configure additional enhancements for SPICE.",
707 type
=> 'string', format
=> 'pve-tag-list',
708 description
=> 'Tags of the VM. This is only meta information.',
714 description
=> "Configure a VirtIO-based Random Number Generator.",
719 format
=> $meta_info_fmt,
720 description
=> "Some (read-only) meta-information about this guest.",
724 type
=> 'string', format
=> 'pve-cpuset',
725 description
=> "List of host cores used to execute guest processes, for example: 0,5,8-11",
734 description
=> 'Specify a custom file containing all meta data passed to the VM via"
735 ." cloud-init. This is provider specific meaning configdrive2 and nocloud differ.',
736 format
=> 'pve-volume-id',
737 format_description
=> 'volume',
742 description
=> 'To pass a custom file containing all network data to the VM via cloud-init.',
743 format
=> 'pve-volume-id',
744 format_description
=> 'volume',
749 description
=> 'To pass a custom file containing all user data to the VM via cloud-init.',
750 format
=> 'pve-volume-id',
751 format_description
=> 'volume',
756 description
=> 'To pass a custom file containing all vendor data to the VM via cloud-init.',
757 format
=> 'pve-volume-id',
758 format_description
=> 'volume',
761 PVE
::JSONSchema
::register_format
('pve-qm-cicustom', $cicustom_fmt);
763 my $confdesc_cloudinit = {
767 description
=> 'Specifies the cloud-init configuration format. The default depends on the'
768 .' configured operating system type (`ostype`. We use the `nocloud` format for Linux,'
769 .' and `configdrive2` for windows.',
770 enum
=> ['configdrive2', 'nocloud', 'opennebula'],
775 description
=> "cloud-init: User name to change ssh keys and password for instead of the"
776 ." image's configured default user.",
781 description
=> 'cloud-init: Password to assign the user. Using this is generally not'
782 .' recommended. Use ssh keys instead. Also note that older cloud-init versions do not'
783 .' support hashed passwords.',
788 description
=> 'cloud-init: do an automatic package upgrade after the first boot.',
794 description
=> 'cloud-init: Specify custom files to replace the automatically generated'
796 format
=> 'pve-qm-cicustom',
801 description
=> 'cloud-init: Sets DNS search domains for a container. Create will'
802 .' automatically use the setting from the host if neither searchdomain nor nameserver'
807 type
=> 'string', format
=> 'address-list',
808 description
=> 'cloud-init: Sets DNS server IP address for a container. Create will'
809 .' automatically use the setting from the host if neither searchdomain nor nameserver'
815 format
=> 'urlencoded',
816 description
=> "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
820 # what about other qemu settings ?
822 #machine => 'string',
835 ##soundhw => 'string',
837 while (my ($k, $v) = each %$confdesc) {
838 PVE
::JSONSchema
::register_standard_option
("pve-qm-$k", $v);
842 my $MAX_SERIAL_PORTS = 4;
843 my $MAX_PARALLEL_PORTS = 3;
849 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
850 description
=> "CPUs accessing this NUMA node.",
851 format_description
=> "id[-id];...",
855 description
=> "Amount of memory this NUMA node provides.",
860 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
861 description
=> "Host NUMA nodes to use.",
862 format_description
=> "id[-id];...",
867 enum
=> [qw(preferred bind interleave)],
868 description
=> "NUMA allocation policy.",
872 PVE
::JSONSchema
::register_format
('pve-qm-numanode', $numa_fmt);
875 type
=> 'string', format
=> $numa_fmt,
876 description
=> "NUMA topology.",
878 PVE
::JSONSchema
::register_standard_option
("pve-qm-numanode", $numadesc);
880 for (my $i = 0; $i < $MAX_NUMA; $i++) {
881 $confdesc->{"numa$i"} = $numadesc;
884 my $nic_model_list = [
900 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
902 my $net_fmt_bridge_descr = <<__EOD__;
903 Bridge to attach the network device to. The Proxmox VE standard bridge
906 If you do not specify a bridge, we create a kvm user (NATed) network
907 device, which provides DHCP and DNS services. The following addresses
914 The DHCP server assign addresses to the guest starting from 10.0.2.15.
918 macaddr
=> get_standard_option
('mac-addr', {
919 description
=> "MAC address. That address must be unique withing your network. This is"
920 ." automatically generated if not specified.",
924 description
=> "Network Card Model. The 'virtio' model provides the best performance with"
925 ." very low CPU overhead. If your guest does not support this driver, it is usually"
926 ." best to use 'e1000'.",
927 enum
=> $nic_model_list,
930 (map { $_ => { keyAlias
=> 'model', alias
=> 'macaddr' }} @$nic_model_list),
931 bridge
=> get_standard_option
('pve-bridge-id', {
932 description
=> $net_fmt_bridge_descr,
937 minimum
=> 0, maximum
=> 64,
938 description
=> 'Number of packet queues to be used on the device.',
944 description
=> "Rate limit in mbps (megabytes per second) as floating point number.",
949 minimum
=> 1, maximum
=> 4094,
950 description
=> 'VLAN tag to apply to packets on this interface.',
955 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
956 description
=> 'VLAN trunks to pass through this interface.',
957 format_description
=> 'vlanid[;vlanid...]',
962 description
=> 'Whether this interface should be protected by the firewall.',
967 description
=> 'Whether this interface should be disconnected (like pulling the plug).',
972 minimum
=> 1, maximum
=> 65520,
973 description
=> "Force MTU, for VirtIO only. Set to '1' to use the bridge MTU",
980 type
=> 'string', format
=> $net_fmt,
981 description
=> "Specify network devices.",
984 PVE
::JSONSchema
::register_standard_option
("pve-qm-net", $netdesc);
989 format
=> 'pve-ipv4-config',
990 format_description
=> 'IPv4Format/CIDR',
991 description
=> 'IPv4 address in CIDR format.',
998 format_description
=> 'GatewayIPv4',
999 description
=> 'Default gateway for IPv4 traffic.',
1005 format
=> 'pve-ipv6-config',
1006 format_description
=> 'IPv6Format/CIDR',
1007 description
=> 'IPv6 address in CIDR format.',
1014 format_description
=> 'GatewayIPv6',
1015 description
=> 'Default gateway for IPv6 traffic.',
1020 PVE
::JSONSchema
::register_format
('pve-qm-ipconfig', $ipconfig_fmt);
1021 my $ipconfigdesc = {
1023 type
=> 'string', format
=> 'pve-qm-ipconfig',
1024 description
=> <<'EODESCR',
1025 cloud-init: Specify IP addresses and gateways for the corresponding interface.
1027 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
1029 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit
1030 gateway should be provided.
1031 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration. This requires
1032 cloud-init 19.4 or newer.
1034 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using
1038 PVE
::JSONSchema
::register_standard_option
("pve-qm-ipconfig", $netdesc);
1040 for (my $i = 0; $i < $MAX_NETS; $i++) {
1041 $confdesc->{"net$i"} = $netdesc;
1042 $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
1045 foreach my $key (keys %$confdesc_cloudinit) {
1046 $confdesc->{$key} = $confdesc_cloudinit->{$key};
1049 PVE
::JSONSchema
::register_format
('pve-cpuset', \
&pve_verify_cpuset
);
1050 sub pve_verify_cpuset
{
1051 my ($set_text, $noerr) = @_;
1053 my ($count, $members) = eval { PVE
::CpuSet
::parse_cpuset
($set_text) };
1057 die "unable to parse cpuset option\n";
1060 return PVE
::CpuSet-
>new($members)->short_string();
1063 PVE
::JSONSchema
::register_format
('pve-volume-id-or-qm-path', \
&verify_volume_id_or_qm_path
);
1064 sub verify_volume_id_or_qm_path
{
1065 my ($volid, $noerr) = @_;
1067 return $volid if $volid eq 'none' || $volid eq 'cdrom';
1069 return verify_volume_id_or_absolute_path
($volid, $noerr);
1072 PVE
::JSONSchema
::register_format
('pve-volume-id-or-absolute-path', \
&verify_volume_id_or_absolute_path
);
1073 sub verify_volume_id_or_absolute_path
{
1074 my ($volid, $noerr) = @_;
1076 return $volid if $volid =~ m
|^/|;
1078 $volid = eval { PVE
::JSONSchema
::check_format
('pve-volume-id', $volid, '') };
1089 pattern
=> '(/dev/.+|socket)',
1090 description
=> "Create a serial device inside the VM (n is 0 to 3)",
1091 verbose_description
=> <<EODESCR,
1092 Create a serial device inside the VM (n is 0 to 3), and pass through a
1093 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1094 host side (use 'qm terminal' to open a terminal connection).
1096 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines -
1097 use with special care.
1099 CAUTION: Experimental! User reported problems with this option.
1106 pattern
=> '/dev/parport\d+|/dev/usb/lp\d+',
1107 description
=> "Map host parallel devices (n is 0 to 2).",
1108 verbose_description
=> <<EODESCR,
1109 Map host parallel devices (n is 0 to 2).
1111 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1112 machines - use with special care.
1114 CAUTION: Experimental! User reported problems with this option.
1118 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
1119 $confdesc->{"parallel$i"} = $paralleldesc;
1122 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
1123 $confdesc->{"serial$i"} = $serialdesc;
1126 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
1127 $confdesc->{"hostpci$i"} = $PVE::QemuServer
::PCI
::hostpcidesc
;
1130 for my $key (keys %{$PVE::QemuServer
::Drive
::drivedesc_hash
}) {
1131 $confdesc->{$key} = $PVE::QemuServer
::Drive
::drivedesc_hash-
>{$key};
1134 for (my $i = 0; $i < $PVE::QemuServer
::USB
::MAX_USB_DEVICES
; $i++) {
1135 $confdesc->{"usb$i"} = $PVE::QemuServer
::USB
::usbdesc
;
1143 description
=> "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n)."
1144 . " Deprecated, use 'order=' instead.",
1145 pattern
=> '[acdn]{1,4}',
1146 format_description
=> "[acdn]{1,4}",
1148 # note: this is also the fallback if boot: is not given at all
1154 format
=> 'pve-qm-bootdev-list',
1155 format_description
=> "device[;device...]",
1156 description
=> <<EODESC,
1157 The guest will attempt to boot from devices in the order they appear here.
1159 Disks, optical drives and passed-through storage USB devices will be directly
1160 booted from, NICs will load PXE, and PCIe devices will either behave like disks
1161 (e.g. NVMe) or load an option ROM (e.g. RAID controller, hardware NIC).
1163 Note that only devices in this list will be marked as bootable and thus loaded
1164 by the guest firmware (BIOS/UEFI). If you require multiple disks for booting
1165 (e.g. software-raid), you need to specify all of them here.
1167 Overrides the deprecated 'legacy=[acdn]*' value when given.
1171 PVE
::JSONSchema
::register_format
('pve-qm-boot', $boot_fmt);
1173 PVE
::JSONSchema
::register_format
('pve-qm-bootdev', \
&verify_bootdev
);
1174 sub verify_bootdev
{
1175 my ($dev, $noerr) = @_;
1177 my $special = $dev =~ m/^efidisk/ || $dev =~ m/^tpmstate/;
1178 return $dev if PVE
::QemuServer
::Drive
::is_valid_drivename
($dev) && !$special;
1182 return 0 if $dev !~ m/^$base\d+$/;
1183 return 0 if !$confdesc->{$dev};
1187 return $dev if $check->("net");
1188 return $dev if $check->("usb");
1189 return $dev if $check->("hostpci");
1192 die "invalid boot device '$dev'\n";
1195 sub print_bootorder
{
1197 return "" if !@$devs;
1198 my $data = { order
=> join(';', @$devs) };
1199 return PVE
::JSONSchema
::print_property_string
($data, $boot_fmt);
1202 my $kvm_api_version = 0;
1205 return $kvm_api_version if $kvm_api_version;
1207 open my $fh, '<', '/dev/kvm' or return;
1209 # 0xae00 => KVM_GET_API_VERSION
1210 $kvm_api_version = ioctl($fh, 0xae00, 0);
1213 return $kvm_api_version;
1216 my $kvm_user_version = {};
1219 sub kvm_user_version
{
1222 $binary //= get_command_for_arch
(get_host_arch
()); # get the native arch by default
1223 my $st = stat($binary);
1225 my $cachedmtime = $kvm_mtime->{$binary} // -1;
1226 return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1227 $cachedmtime == $st->mtime;
1229 $kvm_user_version->{$binary} = 'unknown';
1230 $kvm_mtime->{$binary} = $st->mtime;
1234 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1235 $kvm_user_version->{$binary} = $2;
1239 eval { run_command
([$binary, '--version'], outfunc
=> $code); };
1242 return $kvm_user_version->{$binary};
1245 my sub extract_version
{
1246 my ($machine_type, $version) = @_;
1247 $version = kvm_user_version
() if !defined($version);
1248 return PVE
::QemuServer
::Machine
::extract_version
($machine_type, $version)
1251 sub kernel_has_vhost_net
{
1252 return -c
'/dev/vhost-net';
1257 return defined($confdesc->{$key});
1261 sub get_cdrom_path
{
1263 return $cdrom_path if defined($cdrom_path);
1265 $cdrom_path = first
{ -l
$_ } map { "/dev/cdrom$_" } ('', '1', '2');
1267 if (!defined($cdrom_path)) {
1268 log_warn
("no physical CD-ROM available, ignoring");
1276 my ($storecfg, $vmid, $cdrom) = @_;
1278 if ($cdrom eq 'cdrom') {
1279 return get_cdrom_path
();
1280 } elsif ($cdrom eq 'none') {
1282 } elsif ($cdrom =~ m
|^/|) {
1285 return PVE
::Storage
::path
($storecfg, $cdrom);
1289 # try to convert old style file names to volume IDs
1290 sub filename_to_volume_id
{
1291 my ($vmid, $file, $media) = @_;
1293 if (!($file eq 'none' || $file eq 'cdrom' ||
1294 $file =~ m
|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1296 return if $file =~ m
|/|;
1298 if ($media && $media eq 'cdrom') {
1299 $file = "local:iso/$file";
1301 $file = "local:$vmid/$file";
1308 sub verify_media_type
{
1309 my ($opt, $vtype, $media) = @_;
1314 if ($media eq 'disk') {
1316 } elsif ($media eq 'cdrom') {
1319 die "internal error";
1322 return if ($vtype eq $etype);
1324 raise_param_exc
({ $opt => "unexpected media type ($vtype != $etype)" });
1327 sub cleanup_drive_path
{
1328 my ($opt, $storecfg, $drive) = @_;
1330 # try to convert filesystem paths to volume IDs
1332 if (($drive->{file
} !~ m/^(cdrom|none)$/) &&
1333 ($drive->{file
} !~ m
|^/dev/.+|) &&
1334 ($drive->{file
} !~ m/^([^:]+):(.+)$/) &&
1335 ($drive->{file
} !~ m/^\d+$/)) {
1336 my ($vtype, $volid) = PVE
::Storage
::path_to_volume_id
($storecfg, $drive->{file
});
1337 raise_param_exc
({ $opt => "unable to associate path '$drive->{file}' to any storage"})
1339 $drive->{media
} = 'cdrom' if !$drive->{media
} && $vtype eq 'iso';
1340 verify_media_type
($opt, $vtype, $drive->{media
});
1341 $drive->{file
} = $volid;
1344 $drive->{media
} = 'cdrom' if !$drive->{media
} && $drive->{file
} =~ m/^(cdrom|none)$/;
1347 sub parse_hotplug_features
{
1352 return $res if $data eq '0';
1354 $data = $confdesc->{hotplug
}->{default} if $data eq '1';
1356 foreach my $feature (PVE
::Tools
::split_list
($data)) {
1357 if ($feature =~ m/^(network|disk|cpu|memory|usb|cloudinit)$/) {
1360 die "invalid hotplug feature '$feature'\n";
1366 PVE
::JSONSchema
::register_format
('pve-hotplug-features', \
&pve_verify_hotplug_features
);
1367 sub pve_verify_hotplug_features
{
1368 my ($value, $noerr) = @_;
1370 return $value if parse_hotplug_features
($value);
1374 die "unable to parse hotplug option\n";
1378 my($fh, $noerr) = @_;
1381 my $SG_GET_VERSION_NUM = 0x2282;
1383 my $versionbuf = "\x00" x
8;
1384 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1386 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1389 my $version = unpack("I", $versionbuf);
1390 if ($version < 30000) {
1391 die "scsi generic interface too old\n" if !$noerr;
1395 my $buf = "\x00" x
36;
1396 my $sensebuf = "\x00" x
8;
1397 my $cmd = pack("C x3 C x1", 0x12, 36);
1399 # see /usr/include/scsi/sg.h
1400 my $sg_io_hdr_t = "i i C C s I P P P I I i P C C C C S S i I I";
1403 $sg_io_hdr_t, ord('S'), -3, length($cmd), length($sensebuf), 0, length($buf), $buf, $cmd, $sensebuf, 6000
1406 $ret = ioctl($fh, $SG_IO, $packet);
1408 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1412 my @res = unpack($sg_io_hdr_t, $packet);
1413 if ($res[17] || $res[18]) {
1414 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1419 $res->@{qw(type removable vendor product revision)} = unpack("C C x6 A8 A16 A4", $buf);
1421 $res->{removable
} = $res->{removable
} & 128 ?
1 : 0;
1422 $res->{type
} &= 0x1F;
1430 my $fh = IO
::File-
>new("+<$path") || return;
1431 my $res = scsi_inquiry
($fh, 1);
1437 sub print_tabletdevice_full
{
1438 my ($conf, $arch) = @_;
1440 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1442 # we use uhci for old VMs because tablet driver was buggy in older qemu
1444 if ($q35 || $arch eq 'aarch64') {
1450 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1453 sub print_keyboarddevice_full
{
1454 my ($conf, $arch) = @_;
1456 return if $arch ne 'aarch64';
1458 return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1461 my sub get_drive_id
{
1463 return "$drive->{interface}$drive->{index}";
1466 sub print_drivedevice_full
{
1467 my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1472 my $drive_id = get_drive_id
($drive);
1473 if ($drive->{interface
} eq 'virtio') {
1474 my $pciaddr = print_pci_addr
("$drive_id", $bridges, $arch, $machine_type);
1475 $device = "virtio-blk-pci,drive=drive-$drive_id,id=${drive_id}${pciaddr}";
1476 $device .= ",iothread=iothread-$drive_id" if $drive->{iothread
};
1477 } elsif ($drive->{interface
} eq 'scsi') {
1479 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
1480 my $unit = $drive->{index} % $maxdev;
1481 my $devicetype = 'hd';
1483 if (drive_is_cdrom
($drive)) {
1486 if ($drive->{file
} =~ m
|^/|) {
1487 $path = $drive->{file
};
1488 if (my $info = path_is_scsi
($path)) {
1489 if ($info->{type
} == 0 && $drive->{scsiblock
}) {
1490 $devicetype = 'block';
1491 } elsif ($info->{type
} == 1) { # tape
1492 $devicetype = 'generic';
1496 $path = PVE
::Storage
::path
($storecfg, $drive->{file
});
1499 # for compatibility only, we prefer scsi-hd (#2408, #2355, #2380)
1500 my $version = extract_version
($machine_type, kvm_user_version
());
1501 if ($path =~ m/^iscsi\:\/\
// &&
1502 !min_version
($version, 4, 1)) {
1503 $devicetype = 'generic';
1507 if (!$conf->{scsihw
} || $conf->{scsihw
} =~ m/^lsi/ || $conf->{scsihw
} eq 'pvscsi') {
1508 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit";
1510 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0"
1511 .",lun=$drive->{index}";
1513 $device .= ",drive=drive-$drive_id,id=$drive_id";
1515 if ($drive->{ssd
} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1516 $device .= ",rotation_rate=1";
1518 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1520 } elsif ($drive->{interface
} eq 'ide' || $drive->{interface
} eq 'sata') {
1521 my $maxdev = ($drive->{interface
} eq 'sata') ?
$PVE::QemuServer
::Drive
::MAX_SATA_DISKS
: 2;
1522 my $controller = int($drive->{index} / $maxdev);
1523 my $unit = $drive->{index} % $maxdev;
1524 my $devicetype = ($drive->{media
} && $drive->{media
} eq 'cdrom') ?
"cd" : "hd";
1526 $device = "ide-$devicetype";
1527 if ($drive->{interface
} eq 'ide') {
1528 $device .= ",bus=ide.$controller,unit=$unit";
1530 $device .= ",bus=ahci$controller.$unit";
1532 $device .= ",drive=drive-$drive_id,id=$drive_id";
1534 if ($devicetype eq 'hd') {
1535 if (my $model = $drive->{model
}) {
1536 $model = URI
::Escape
::uri_unescape
($model);
1537 $device .= ",model=$model";
1539 if ($drive->{ssd
}) {
1540 $device .= ",rotation_rate=1";
1543 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1544 } elsif ($drive->{interface
} eq 'usb') {
1546 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1548 die "unsupported interface type";
1551 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex
};
1553 if (my $serial = $drive->{serial
}) {
1554 $serial = URI
::Escape
::uri_unescape
($serial);
1555 $device .= ",serial=$serial";
1562 sub get_initiator_name
{
1565 my $fh = IO
::File-
>new('/etc/iscsi/initiatorname.iscsi') || return;
1566 while (defined(my $line = <$fh>)) {
1567 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1576 my sub storage_allows_io_uring_default
{
1577 my ($scfg, $cache_direct) = @_;
1579 # io_uring with cache mode writeback or writethrough on krbd will hang...
1580 return if $scfg && $scfg->{type
} eq 'rbd' && $scfg->{krbd
} && !$cache_direct;
1582 # io_uring with cache mode writeback or writethrough on LVM will hang, without cache only
1583 # sometimes, just plain disable...
1584 return if $scfg && $scfg->{type
} eq 'lvm';
1586 # io_uring causes problems when used with CIFS since kernel 5.15
1587 # Some discussion: https://www.spinics.net/lists/linux-cifs/msg26734.html
1588 return if $scfg && $scfg->{type
} eq 'cifs';
1593 my sub drive_uses_cache_direct
{
1594 my ($drive, $scfg) = @_;
1596 my $cache_direct = 0;
1598 if (my $cache = $drive->{cache
}) {
1599 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1600 } elsif (!drive_is_cdrom
($drive) && !($scfg && $scfg->{type
} eq 'btrfs' && !$scfg->{nocow
})) {
1604 return $cache_direct;
1607 sub print_drive_commandline_full
{
1608 my ($storecfg, $vmid, $drive, $pbs_name, $io_uring) = @_;
1611 my $volid = $drive->{file
};
1612 my $format = $drive->{format
};
1613 my $drive_id = get_drive_id
($drive);
1615 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1616 my $scfg = $storeid ? PVE
::Storage
::storage_config
($storecfg, $storeid) : undef;
1618 if (drive_is_cdrom
($drive)) {
1619 $path = get_iso_path
($storecfg, $vmid, $volid);
1620 die "$drive_id: cannot back cdrom drive with PBS snapshot\n" if $pbs_name;
1623 $path = PVE
::Storage
::path
($storecfg, $volid);
1624 $format //= qemu_img_format
($scfg, $volname);
1631 my $is_rbd = $path =~ m/^rbd:/;
1634 my @qemu_drive_options = qw(heads secs cyls trans media cache rerror werror aio discard);
1635 foreach my $o (@qemu_drive_options) {
1636 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1639 # snapshot only accepts on|off
1640 if (defined($drive->{snapshot
})) {
1641 my $v = $drive->{snapshot
} ?
'on' : 'off';
1642 $opts .= ",snapshot=$v";
1645 if (defined($drive->{ro
})) { # ro maps to QEMUs `readonly`, which accepts `on` or `off` only
1646 $opts .= ",readonly=" . ($drive->{ro
} ?
'on' : 'off');
1649 foreach my $type (['', '-total'], [_rd
=> '-read'], [_wr
=> '-write']) {
1650 my ($dir, $qmpname) = @$type;
1651 if (my $v = $drive->{"mbps$dir"}) {
1652 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1654 if (my $v = $drive->{"mbps${dir}_max"}) {
1655 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1657 if (my $v = $drive->{"bps${dir}_max_length"}) {
1658 $opts .= ",throttling.bps$qmpname-max-length=$v";
1660 if (my $v = $drive->{"iops${dir}"}) {
1661 $opts .= ",throttling.iops$qmpname=$v";
1663 if (my $v = $drive->{"iops${dir}_max"}) {
1664 $opts .= ",throttling.iops$qmpname-max=$v";
1666 if (my $v = $drive->{"iops${dir}_max_length"}) {
1667 $opts .= ",throttling.iops$qmpname-max-length=$v";
1672 $format = "rbd" if $is_rbd;
1673 die "$drive_id: Proxmox Backup Server backed drive cannot auto-detect the format\n"
1675 $opts .= ",format=alloc-track,file.driver=$format";
1677 $opts .= ",format=$format";
1680 my $cache_direct = drive_uses_cache_direct
($drive, $scfg);
1682 $opts .= ",cache=none" if !$drive->{cache
} && $cache_direct;
1684 if (!$drive->{aio
}) {
1685 if ($io_uring && storage_allows_io_uring_default
($scfg, $cache_direct)) {
1686 # io_uring supports all cache modes
1687 $opts .= ",aio=io_uring";
1689 # aio native works only with O_DIRECT
1691 $opts .= ",aio=native";
1693 $opts .= ",aio=threads";
1698 if (!drive_is_cdrom
($drive)) {
1700 if (defined($drive->{detect_zeroes
}) && !$drive->{detect_zeroes
}) {
1701 $detectzeroes = 'off';
1702 } elsif ($drive->{discard
}) {
1703 $detectzeroes = $drive->{discard
} eq 'on' ?
'unmap' : 'on';
1705 # This used to be our default with discard not being specified:
1706 $detectzeroes = 'on';
1709 # note: 'detect-zeroes' works per blockdev and we want it to persist
1710 # after the alloc-track is removed, so put it on 'file' directly
1711 my $dz_param = $pbs_name ?
"file.detect-zeroes" : "detect-zeroes";
1712 $opts .= ",$dz_param=$detectzeroes" if $detectzeroes;
1716 $opts .= ",backing=$pbs_name";
1717 $opts .= ",auto-remove=on";
1720 # my $file_param = $pbs_name ? "file.file.filename" : "file";
1721 my $file_param = "file";
1723 # non-rbd drivers require the underlying file to be a seperate block
1724 # node, so add a second .file indirection
1725 $file_param .= ".file" if !$is_rbd;
1726 $file_param .= ".filename";
1728 my $pathinfo = $path ?
"$file_param=$path," : '';
1730 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1733 sub print_pbs_blockdev
{
1734 my ($pbs_conf, $pbs_name) = @_;
1735 my $blockdev = "driver=pbs,node-name=$pbs_name,read-only=on";
1736 $blockdev .= ",repository=$pbs_conf->{repository}";
1737 $blockdev .= ",namespace=$pbs_conf->{namespace}" if $pbs_conf->{namespace
};
1738 $blockdev .= ",snapshot=$pbs_conf->{snapshot}";
1739 $blockdev .= ",archive=$pbs_conf->{archive}";
1740 $blockdev .= ",keyfile=$pbs_conf->{keyfile}" if $pbs_conf->{keyfile
};
1744 sub print_netdevice_full
{
1745 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version) = @_;
1747 my $device = $net->{model
};
1748 if ($net->{model
} eq 'virtio') {
1749 $device = 'virtio-net-pci';
1752 my $pciaddr = print_pci_addr
("$netid", $bridges, $arch, $machine_type);
1753 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1754 if ($net->{queues
} && $net->{queues
} > 1 && $net->{model
} eq 'virtio'){
1755 # Consider we have N queues, the number of vectors needed is 2 * N + 2, i.e., one per in
1756 # and out of each queue plus one config interrupt and control vector queue
1757 my $vectors = $net->{queues
} * 2 + 2;
1758 $tmpstr .= ",vectors=$vectors,mq=on";
1759 if (min_version
($machine_version, 7, 1)) {
1760 $tmpstr .= ",packed=on";
1764 if (min_version
($machine_version, 7, 1) && $net->{model
} eq 'virtio'){
1765 $tmpstr .= ",rx_queue_size=1024,tx_queue_size=1024";
1768 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex
} ;
1770 if (my $mtu = $net->{mtu
}) {
1771 if ($net->{model
} eq 'virtio' && $net->{bridge
}) {
1772 my $bridge_mtu = PVE
::Network
::read_bridge_mtu
($net->{bridge
});
1775 } elsif ($mtu < 576) {
1776 die "netdev $netid: MTU '$mtu' is smaller than the IP minimum MTU '576'\n";
1777 } elsif ($mtu > $bridge_mtu) {
1778 die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
1780 $tmpstr .= ",host_mtu=$mtu";
1782 warn "WARN: netdev $netid: ignoring MTU '$mtu', not using VirtIO or no bridge configured.\n";
1786 if ($use_old_bios_files) {
1788 if ($device eq 'virtio-net-pci') {
1789 $romfile = 'pxe-virtio.rom';
1790 } elsif ($device eq 'e1000') {
1791 $romfile = 'pxe-e1000.rom';
1792 } elsif ($device eq 'e1000e') {
1793 $romfile = 'pxe-e1000e.rom';
1794 } elsif ($device eq 'ne2k') {
1795 $romfile = 'pxe-ne2k_pci.rom';
1796 } elsif ($device eq 'pcnet') {
1797 $romfile = 'pxe-pcnet.rom';
1798 } elsif ($device eq 'rtl8139') {
1799 $romfile = 'pxe-rtl8139.rom';
1801 $tmpstr .= ",romfile=$romfile" if $romfile;
1807 sub print_netdev_full
{
1808 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
1811 if ($netid =~ m/^net(\d+)$/) {
1815 die "got strange net id '$i'\n" if $i >= ${MAX_NETS
};
1817 my $ifname = "tap${vmid}i$i";
1819 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1820 die "interface name '$ifname' is too long (max 15 character)\n"
1821 if length($ifname) >= 16;
1823 my $vhostparam = '';
1824 if (is_native
($arch)) {
1825 $vhostparam = ',vhost=on' if kernel_has_vhost_net
() && $net->{model
} eq 'virtio';
1828 my $vmname = $conf->{name
} || "vm$vmid";
1831 my $script = $hotplug ?
"pve-bridge-hotplug" : "pve-bridge";
1833 if ($net->{bridge
}) {
1834 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script"
1835 .",downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1837 $netdev = "type=user,id=$netid,hostname=$vmname";
1840 $netdev .= ",queues=$net->{queues}" if ($net->{queues
} && $net->{model
} eq 'virtio');
1846 'cirrus' => 'cirrus-vga',
1848 'vmware' => 'vmware-svga',
1849 'virtio' => 'virtio-vga',
1850 'virtio-gl' => 'virtio-vga-gl',
1853 sub print_vga_device
{
1854 my ($conf, $vga, $arch, $machine_version, $machine, $id, $qxlnum, $bridges) = @_;
1856 my $type = $vga_map->{$vga->{type
}};
1857 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
1858 $type = 'virtio-gpu';
1860 my $vgamem_mb = $vga->{memory
};
1862 my $max_outputs = '';
1864 $type = $id ?
'qxl' : 'qxl-vga';
1866 if (!$conf->{ostype
} || $conf->{ostype
} =~ m/^(?:l\d\d)|(?:other)$/) {
1867 # set max outputs so linux can have up to 4 qxl displays with one device
1868 if (min_version
($machine_version, 4, 1)) {
1869 $max_outputs = ",max_outputs=4";
1874 die "no devicetype for $vga->{type}\n" if !$type;
1878 if ($vga->{type
} =~ /^virtio/) {
1879 my $bytes = PVE
::Tools
::convert_size
($vgamem_mb, "mb" => "b");
1880 $memory = ",max_hostmem=$bytes";
1882 # from https://www.spice-space.org/multiple-monitors.html
1883 $memory = ",vgamem_mb=$vga->{memory}";
1884 my $ram = $vgamem_mb * 4;
1885 my $vram = $vgamem_mb * 2;
1886 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
1888 $memory = ",vgamem_mb=$vga->{memory}";
1890 } elsif ($qxlnum && $id) {
1891 $memory = ",ram_size=67108864,vram_size=33554432";
1895 if ($type eq 'VGA' && windows_version
($conf->{ostype
})) {
1896 $edidoff=",edid=off" if (!defined($conf->{bios
}) || $conf->{bios
} ne 'ovmf');
1899 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1900 my $vgaid = "vga" . ($id // '');
1902 if ($q35 && $vgaid eq 'vga') {
1903 # the first display uses pcie.0 bus on q35 machines
1904 $pciaddr = print_pcie_addr
($vgaid, $bridges, $arch, $machine);
1906 $pciaddr = print_pci_addr
($vgaid, $bridges, $arch, $machine);
1909 if ($vga->{type
} eq 'virtio-gl') {
1910 my $base = '/usr/lib/x86_64-linux-gnu/lib';
1911 die "missing libraries for '$vga->{type}' detected! Please install 'libgl1' and 'libegl1'\n"
1912 if !-e
"${base}EGL.so.1" || !-e
"${base}GL.so.1";
1914 die "no DRM render node detected (/dev/dri/renderD*), no GPU? - needed for '$vga->{type}' display\n"
1915 if !PVE
::Tools
::dir_glob_regex
('/dev/dri/', "renderD.*");
1918 return "$type,id=${vgaid}${memory}${max_outputs}${pciaddr}${edidoff}";
1921 sub parse_number_sets
{
1924 foreach my $part (split(/;/, $set)) {
1925 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1926 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1927 push @$res, [ $1, $2 ];
1929 die "invalid range: $part\n";
1938 my $res = parse_property_string
($numa_fmt, $data);
1939 $res->{cpus
} = parse_number_sets
($res->{cpus
}) if defined($res->{cpus
});
1940 $res->{hostnodes
} = parse_number_sets
($res->{hostnodes
}) if defined($res->{hostnodes
});
1944 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1946 my ($data, $disable_mac_autogen) = @_;
1948 my $res = eval { parse_property_string
($net_fmt, $data) };
1953 if (!defined($res->{macaddr
}) && !$disable_mac_autogen) {
1954 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
1955 $res->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
});
1960 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
1961 sub parse_ipconfig
{
1964 my $res = eval { parse_property_string
($ipconfig_fmt, $data) };
1970 if ($res->{gw
} && !$res->{ip
}) {
1971 warn 'gateway specified without specifying an IP address';
1974 if ($res->{gw6
} && !$res->{ip6
}) {
1975 warn 'IPv6 gateway specified without specifying an IPv6 address';
1978 if ($res->{gw
} && $res->{ip
} eq 'dhcp') {
1979 warn 'gateway specified together with DHCP';
1982 if ($res->{gw6
} && $res->{ip6
} !~ /^$IPV6RE/) {
1984 warn "IPv6 gateway specified together with $res->{ip6} address";
1988 if (!$res->{ip
} && !$res->{ip6
}) {
1989 return { ip
=> 'dhcp', ip6
=> 'dhcp' };
1998 return PVE
::JSONSchema
::print_property_string
($net, $net_fmt);
2001 sub add_random_macs
{
2002 my ($settings) = @_;
2004 foreach my $opt (keys %$settings) {
2005 next if $opt !~ m/^net(\d+)$/;
2006 my $net = parse_net
($settings->{$opt});
2008 $settings->{$opt} = print_net
($net);
2012 sub vm_is_volid_owner
{
2013 my ($storecfg, $vmid, $volid) = @_;
2015 if ($volid !~ m
|^/|) {
2017 eval { ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid); };
2018 if ($owner && ($owner == $vmid)) {
2026 sub vmconfig_register_unused_drive
{
2027 my ($storecfg, $vmid, $conf, $drive) = @_;
2029 if (drive_is_cloudinit
($drive)) {
2030 eval { PVE
::Storage
::vdisk_free
($storecfg, $drive->{file
}) };
2032 delete $conf->{cloudinit
};
2033 } elsif (!drive_is_cdrom
($drive)) {
2034 my $volid = $drive->{file
};
2035 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
2036 PVE
::QemuConfig-
>add_unused_volume($conf, $volid, $vmid);
2041 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2045 pattern
=> '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2046 format_description
=> 'UUID',
2047 description
=> "Set SMBIOS1 UUID.",
2052 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2053 format_description
=> 'Base64 encoded string',
2054 description
=> "Set SMBIOS1 version.",
2059 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2060 format_description
=> 'Base64 encoded string',
2061 description
=> "Set SMBIOS1 serial number.",
2066 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2067 format_description
=> 'Base64 encoded string',
2068 description
=> "Set SMBIOS1 manufacturer.",
2073 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2074 format_description
=> 'Base64 encoded string',
2075 description
=> "Set SMBIOS1 product ID.",
2080 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2081 format_description
=> 'Base64 encoded string',
2082 description
=> "Set SMBIOS1 SKU string.",
2087 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2088 format_description
=> 'Base64 encoded string',
2089 description
=> "Set SMBIOS1 family string.",
2094 description
=> 'Flag to indicate that the SMBIOS values are base64 encoded',
2102 my $res = eval { parse_property_string
($smbios1_fmt, $data) };
2109 return PVE
::JSONSchema
::print_property_string
($smbios1, $smbios1_fmt);
2112 PVE
::JSONSchema
::register_format
('pve-qm-smbios1', $smbios1_fmt);
2114 sub parse_watchdog
{
2119 my $res = eval { parse_property_string
($watchdog_fmt, $value) };
2124 sub parse_guest_agent
{
2127 return {} if !defined($conf->{agent
});
2129 my $res = eval { parse_property_string
($agent_fmt, $conf->{agent
}) };
2132 # if the agent is disabled ignore the other potentially set properties
2133 return {} if !$res->{enabled
};
2138 my ($conf, $key) = @_;
2139 return undef if !defined($conf->{agent
});
2141 my $agent = parse_guest_agent
($conf);
2142 return $agent->{$key};
2148 return {} if !$value;
2149 my $res = eval { parse_property_string
($vga_fmt, $value) };
2159 my $res = eval { parse_property_string
($rng_fmt, $value) };
2164 sub parse_meta_info
{
2169 my $res = eval { parse_property_string
($meta_info_fmt, $value) };
2174 sub new_meta_info_string
{
2175 my () = @_; # for now do not allow to override any value
2177 return PVE
::JSONSchema
::print_property_string
(
2179 'creation-qemu' => kvm_user_version
(),
2180 ctime
=> "". int(time()),
2186 sub qemu_created_version_fixups
{
2187 my ($conf, $forcemachine, $kvmver) = @_;
2189 my $meta = parse_meta_info
($conf->{meta
}) // {};
2190 my $forced_vers = PVE
::QemuServer
::Machine
::extract_version
($forcemachine);
2192 # check if we need to apply some handling for VMs that always use the latest machine version but
2193 # had a machine version transition happen that affected HW such that, e.g., an OS config change
2194 # would be required (we do not want to pin machine version for non-windows OS type)
2196 (!defined($conf->{machine
}) || $conf->{machine
} =~ m/^(?:pc|q35|virt)$/) # non-versioned machine
2197 && (!defined($meta->{'creation-qemu'}) || !min_version
($meta->{'creation-qemu'}, 6, 1)) # created before 6.1
2198 && (!$forced_vers || min_version
($forced_vers, 6, 1)) # handle snapshot-rollback/migrations
2199 && min_version
($kvmver, 6, 1) # only need to apply the change since 6.1
2201 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
2202 if ($q35 && $conf->{ostype
} && $conf->{ostype
} eq 'l26') {
2203 # this changed to default-on in Q 6.1 for q35 machines, it will mess with PCI slot view
2204 # and thus with the predictable interface naming of systemd
2205 return ['-global', 'ICH9-LPC.acpi-pci-hotplug-with-bridge-support=off'];
2211 # add JSON properties for create and set function
2212 sub json_config_properties
{
2213 my ($prop, $with_disk_alloc) = @_;
2215 my $skip_json_config_opts = {
2219 runningmachine
=> 1,
2224 foreach my $opt (keys %$confdesc) {
2225 next if $skip_json_config_opts->{$opt};
2227 if ($with_disk_alloc && is_valid_drivename
($opt)) {
2228 $prop->{$opt} = $PVE::QemuServer
::Drive
::drivedesc_hash_with_alloc-
>{$opt};
2230 $prop->{$opt} = $confdesc->{$opt};
2237 # Properties that we can read from an OVF file
2238 sub json_ovf_properties
{
2241 for my $device (PVE
::QemuServer
::Drive
::valid_drive_names
()) {
2242 $prop->{$device} = {
2244 format
=> 'pve-volume-id-or-absolute-path',
2245 description
=> "Disk image that gets imported to $device",
2252 description
=> "The number of CPU cores.",
2257 description
=> "Amount of RAM for the VM in MB.",
2262 description
=> "Name of the VM.",
2269 # return copy of $confdesc_cloudinit to generate documentation
2270 sub cloudinit_config_properties
{
2272 return dclone
($confdesc_cloudinit);
2275 sub cloudinit_pending_properties
{
2277 map { $_ => 1 } keys $confdesc_cloudinit->%*,
2280 $p->{"net$_"} = 1 for 0..($MAX_NETS-1);
2285 my ($key, $value) = @_;
2287 die "unknown setting '$key'\n" if !$confdesc->{$key};
2289 my $type = $confdesc->{$key}->{type
};
2291 if (!defined($value)) {
2292 die "got undefined value\n";
2295 if ($value =~ m/[\n\r]/) {
2296 die "property contains a line feed\n";
2299 if ($type eq 'boolean') {
2300 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2301 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2302 die "type check ('boolean') failed - got '$value'\n";
2303 } elsif ($type eq 'integer') {
2304 return int($1) if $value =~ m/^(\d+)$/;
2305 die "type check ('integer') failed - got '$value'\n";
2306 } elsif ($type eq 'number') {
2307 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2308 die "type check ('number') failed - got '$value'\n";
2309 } elsif ($type eq 'string') {
2310 if (my $fmt = $confdesc->{$key}->{format
}) {
2311 PVE
::JSONSchema
::check_format
($fmt, $value);
2314 $value =~ s/^\"(.*)\"$/$1/;
2317 die "internal error"
2322 my ($storecfg, $vmid, $skiplock, $replacement_conf, $purge_unreferenced) = @_;
2324 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2326 if (!$skiplock && !PVE
::QemuConfig-
>has_lock($conf, 'suspended')) {
2327 PVE
::QemuConfig-
>check_lock($conf);
2330 if ($conf->{template
}) {
2331 # check if any base image is still used by a linked clone
2332 PVE
::QemuConfig-
>foreach_volume_full($conf, { include_unused
=> 1 }, sub {
2333 my ($ds, $drive) = @_;
2334 return if drive_is_cdrom
($drive);
2336 my $volid = $drive->{file
};
2337 return if !$volid || $volid =~ m
|^/|;
2339 die "base volume '$volid' is still in use by linked cloned\n"
2340 if PVE
::Storage
::volume_is_base_and_used
($storecfg, $volid);
2346 my $remove_owned_drive = sub {
2347 my ($ds, $drive) = @_;
2348 return if drive_is_cdrom
($drive, 1);
2350 my $volid = $drive->{file
};
2351 return if !$volid || $volid =~ m
|^/|;
2352 return if $volids->{$volid};
2354 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
2355 return if !$path || !$owner || ($owner != $vmid);
2357 $volids->{$volid} = 1;
2358 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2359 warn "Could not remove disk '$volid', check manually: $@" if $@;
2362 # only remove disks owned by this VM (referenced in the config)
2363 my $include_opts = {
2364 include_unused
=> 1,
2365 extra_keys
=> ['vmstate'],
2367 PVE
::QemuConfig-
>foreach_volume_full($conf, $include_opts, $remove_owned_drive);
2369 for my $snap (values %{$conf->{snapshots
}}) {
2370 next if !defined($snap->{vmstate
});
2371 my $drive = PVE
::QemuConfig-
>parse_volume('vmstate', $snap->{vmstate
}, 1);
2372 next if !defined($drive);
2373 $remove_owned_drive->('vmstate', $drive);
2376 PVE
::QemuConfig-
>foreach_volume_full($conf->{pending
}, $include_opts, $remove_owned_drive);
2378 if ($purge_unreferenced) { # also remove unreferenced disk
2379 my $vmdisks = PVE
::Storage
::vdisk_list
($storecfg, undef, $vmid, undef, 'images');
2380 PVE
::Storage
::foreach_volid
($vmdisks, sub {
2381 my ($volid, $sid, $volname, $d) = @_;
2382 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2387 if (defined $replacement_conf) {
2388 PVE
::QemuConfig-
>write_config($vmid, $replacement_conf);
2390 PVE
::QemuConfig-
>destroy_config($vmid);
2394 sub parse_vm_config
{
2395 my ($filename, $raw, $strict) = @_;
2397 return if !defined($raw);
2400 digest
=> Digest
::SHA
::sha1_hex
($raw),
2406 my $handle_error = sub {
2416 $filename =~ m
|/qemu-server/(\d
+)\
.conf
$|
2417 || die "got strange filename '$filename'";
2423 my $finish_description = sub {
2424 if (defined($descr)) {
2426 $conf->{description
} = $descr;
2432 my @lines = split(/\n/, $raw);
2433 foreach my $line (@lines) {
2434 next if $line =~ m/^\s*$/;
2436 if ($line =~ m/^\[PENDING\]\s*$/i) {
2437 $section = 'pending';
2438 $finish_description->();
2439 $conf = $res->{$section} = {};
2441 } elsif ($line =~ m/^\[special:cloudinit\]\s*$/i) {
2442 $section = 'cloudinit';
2443 $finish_description->();
2444 $conf = $res->{$section} = {};
2447 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2449 $finish_description->();
2450 $conf = $res->{snapshots
}->{$section} = {};
2454 if ($line =~ m/^\#(.*)$/) {
2455 $descr = '' if !defined($descr);
2456 $descr .= PVE
::Tools
::decode_text
($1) . "\n";
2460 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2461 $descr = '' if !defined($descr);
2462 $descr .= PVE
::Tools
::decode_text
($2);
2463 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2464 $conf->{snapstate
} = $1;
2465 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2468 $conf->{$key} = $value;
2469 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2471 if ($section eq 'pending') {
2472 $conf->{delete} = $value; # we parse this later
2474 $handle_error->("vm $vmid - property 'delete' is only allowed in [PENDING]\n");
2476 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2479 if ($section eq 'cloudinit') {
2480 # ignore validation only used for informative purpose
2481 $conf->{$key} = $value;
2484 eval { $value = check_type
($key, $value); };
2486 $handle_error->("vm $vmid - unable to parse value of '$key' - $@");
2488 $key = 'ide2' if $key eq 'cdrom';
2489 my $fmt = $confdesc->{$key}->{format
};
2490 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2491 my $v = parse_drive
($key, $value);
2492 if (my $volid = filename_to_volume_id
($vmid, $v->{file
}, $v->{media
})) {
2493 $v->{file
} = $volid;
2494 $value = print_drive
($v);
2496 $handle_error->("vm $vmid - unable to parse value of '$key'\n");
2501 $conf->{$key} = $value;
2504 $handle_error->("vm $vmid - unable to parse config: $line\n");
2508 $finish_description->();
2509 delete $res->{snapstate
}; # just to be sure
2514 sub write_vm_config
{
2515 my ($filename, $conf) = @_;
2517 delete $conf->{snapstate
}; # just to be sure
2519 if ($conf->{cdrom
}) {
2520 die "option ide2 conflicts with cdrom\n" if $conf->{ide2
};
2521 $conf->{ide2
} = $conf->{cdrom
};
2522 delete $conf->{cdrom
};
2525 # we do not use 'smp' any longer
2526 if ($conf->{sockets
}) {
2527 delete $conf->{smp
};
2528 } elsif ($conf->{smp
}) {
2529 $conf->{sockets
} = $conf->{smp
};
2530 delete $conf->{cores
};
2531 delete $conf->{smp
};
2534 my $used_volids = {};
2536 my $cleanup_config = sub {
2537 my ($cref, $pending, $snapname) = @_;
2539 foreach my $key (keys %$cref) {
2540 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2541 $key eq 'snapstate' || $key eq 'pending' || $key eq 'cloudinit';
2542 my $value = $cref->{$key};
2543 if ($key eq 'delete') {
2544 die "propertry 'delete' is only allowed in [PENDING]\n"
2546 # fixme: check syntax?
2549 eval { $value = check_type
($key, $value); };
2550 die "unable to parse value of '$key' - $@" if $@;
2552 $cref->{$key} = $value;
2554 if (!$snapname && is_valid_drivename
($key)) {
2555 my $drive = parse_drive
($key, $value);
2556 $used_volids->{$drive->{file
}} = 1 if $drive && $drive->{file
};
2561 &$cleanup_config($conf);
2563 &$cleanup_config($conf->{pending
}, 1);
2565 foreach my $snapname (keys %{$conf->{snapshots
}}) {
2566 die "internal error: snapshot name '$snapname' is forbidden" if lc($snapname) eq 'pending';
2567 &$cleanup_config($conf->{snapshots
}->{$snapname}, undef, $snapname);
2570 # remove 'unusedX' settings if we re-add a volume
2571 foreach my $key (keys %$conf) {
2572 my $value = $conf->{$key};
2573 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2574 delete $conf->{$key};
2578 my $generate_raw_config = sub {
2579 my ($conf, $pending) = @_;
2583 # add description as comment to top of file
2584 if (defined(my $descr = $conf->{description
})) {
2586 foreach my $cl (split(/\n/, $descr)) {
2587 $raw .= '#' . PVE
::Tools
::encode_text
($cl) . "\n";
2590 $raw .= "#\n" if $pending;
2594 foreach my $key (sort keys %$conf) {
2595 next if $key =~ /^(digest|description|pending|cloudinit|snapshots)$/;
2596 $raw .= "$key: $conf->{$key}\n";
2601 my $raw = &$generate_raw_config($conf);
2603 if (scalar(keys %{$conf->{pending
}})){
2604 $raw .= "\n[PENDING]\n";
2605 $raw .= &$generate_raw_config($conf->{pending
}, 1);
2608 if (scalar(keys %{$conf->{cloudinit
}}) && PVE
::QemuConfig-
>has_cloudinit($conf)){
2609 $raw .= "\n[special:cloudinit]\n";
2610 $raw .= &$generate_raw_config($conf->{cloudinit
});
2613 foreach my $snapname (sort keys %{$conf->{snapshots
}}) {
2614 $raw .= "\n[$snapname]\n";
2615 $raw .= &$generate_raw_config($conf->{snapshots
}->{$snapname});
2625 # we use static defaults from our JSON schema configuration
2626 foreach my $key (keys %$confdesc) {
2627 if (defined(my $default = $confdesc->{$key}->{default})) {
2628 $res->{$key} = $default;
2636 my $vmlist = PVE
::Cluster
::get_vmlist
();
2638 return $res if !$vmlist || !$vmlist->{ids
};
2639 my $ids = $vmlist->{ids
};
2640 my $nodename = nodename
();
2642 foreach my $vmid (keys %$ids) {
2643 my $d = $ids->{$vmid};
2644 next if !$d->{node
} || $d->{node
} ne $nodename;
2645 next if !$d->{type
} || $d->{type
} ne 'qemu';
2646 $res->{$vmid}->{exists} = 1;
2651 # test if VM uses local resources (to prevent migration)
2652 sub check_local_resources
{
2653 my ($conf, $noerr) = @_;
2656 my $mapped_res = [];
2658 my $nodelist = PVE
::Cluster
::get_nodelist
();
2659 my $pci_map = PVE
::Mapping
::PCI
::config
();
2660 my $usb_map = PVE
::Mapping
::USB
::config
();
2662 my $missing_mappings_by_node = { map { $_ => [] } @$nodelist };
2664 my $add_missing_mapping = sub {
2665 my ($type, $key, $id) = @_;
2666 for my $node (@$nodelist) {
2668 if ($type eq 'pci') {
2669 $entry = PVE
::Mapping
::PCI
::get_node_mapping
($pci_map, $id, $node);
2670 } elsif ($type eq 'usb') {
2671 $entry = PVE
::Mapping
::USB
::get_node_mapping
($usb_map, $id, $node);
2673 if (!scalar($entry->@*)) {
2674 push @{$missing_mappings_by_node->{$node}}, $key;
2679 push @loc_res, "hostusb" if $conf->{hostusb
}; # old syntax
2680 push @loc_res, "hostpci" if $conf->{hostpci
}; # old syntax
2682 push @loc_res, "ivshmem" if $conf->{ivshmem
};
2684 foreach my $k (keys %$conf) {
2685 if ($k =~ m/^usb/) {
2686 my $entry = parse_property_string
('pve-qm-usb', $conf->{$k});
2687 next if $entry->{host
} =~ m/^spice$/i;
2688 if ($entry->{mapping
}) {
2689 $add_missing_mapping->('usb', $k, $entry->{mapping
});
2690 push @$mapped_res, $k;
2693 if ($k =~ m/^hostpci/) {
2694 my $entry = parse_property_string
('pve-qm-hostpci', $conf->{$k});
2695 if ($entry->{mapping
}) {
2696 $add_missing_mapping->('pci', $k, $entry->{mapping
});
2697 push @$mapped_res, $k;
2700 # sockets are safe: they will recreated be on the target side post-migrate
2701 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2702 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2705 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2707 return wantarray ?
(\
@loc_res, $mapped_res, $missing_mappings_by_node) : \
@loc_res;
2710 # check if used storages are available on all nodes (use by migrate)
2711 sub check_storage_availability
{
2712 my ($storecfg, $conf, $node) = @_;
2714 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2715 my ($ds, $drive) = @_;
2717 my $volid = $drive->{file
};
2720 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2723 # check if storage is available on both nodes
2724 my $scfg = PVE
::Storage
::storage_check_enabled
($storecfg, $sid);
2725 PVE
::Storage
::storage_check_enabled
($storecfg, $sid, $node);
2727 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $volid);
2729 die "$volid: content type '$vtype' is not available on storage '$sid'\n"
2730 if !$scfg->{content
}->{$vtype};
2734 # list nodes where all VM images are available (used by has_feature API)
2736 my ($conf, $storecfg) = @_;
2738 my $nodelist = PVE
::Cluster
::get_nodelist
();
2739 my $nodehash = { map { $_ => 1 } @$nodelist };
2740 my $nodename = nodename
();
2742 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2743 my ($ds, $drive) = @_;
2745 my $volid = $drive->{file
};
2748 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2750 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2751 if ($scfg->{disable
}) {
2753 } elsif (my $avail = $scfg->{nodes
}) {
2754 foreach my $node (keys %$nodehash) {
2755 delete $nodehash->{$node} if !$avail->{$node};
2757 } elsif (!$scfg->{shared
}) {
2758 foreach my $node (keys %$nodehash) {
2759 delete $nodehash->{$node} if $node ne $nodename
2768 sub check_local_storage_availability
{
2769 my ($conf, $storecfg) = @_;
2771 my $nodelist = PVE
::Cluster
::get_nodelist
();
2772 my $nodehash = { map { $_ => {} } @$nodelist };
2774 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2775 my ($ds, $drive) = @_;
2777 my $volid = $drive->{file
};
2780 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2782 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2784 if ($scfg->{disable
}) {
2785 foreach my $node (keys %$nodehash) {
2786 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2788 } elsif (my $avail = $scfg->{nodes
}) {
2789 foreach my $node (keys %$nodehash) {
2790 if (!$avail->{$node}) {
2791 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2798 foreach my $node (values %$nodehash) {
2799 if (my $unavail = $node->{unavailable_storages
}) {
2800 $node->{unavailable_storages
} = [ sort keys %$unavail ];
2807 # Compat only, use assert_config_exists_on_node and vm_running_locally where possible
2809 my ($vmid, $nocheck, $node) = @_;
2811 # $nocheck is set when called during a migration, in which case the config
2812 # file might still or already reside on the *other* node
2813 # - because rename has already happened, and current node is source
2814 # - because rename hasn't happened yet, and current node is target
2815 # - because rename has happened, current node is target, but hasn't yet
2817 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid, $node) if !$nocheck;
2818 return PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
2823 my $vzlist = config_list
();
2825 my $fd = IO
::Dir-
>new($PVE::QemuServer
::Helpers
::var_run_tmpdir
) || return $vzlist;
2827 while (defined(my $de = $fd->read)) {
2828 next if $de !~ m/^(\d+)\.pid$/;
2830 next if !defined($vzlist->{$vmid});
2831 if (my $pid = check_running
($vmid)) {
2832 $vzlist->{$vmid}->{pid
} = $pid;
2839 our $vmstatus_return_properties = {
2840 vmid
=> get_standard_option
('pve-vmid'),
2842 description
=> "QEMU process status.",
2844 enum
=> ['stopped', 'running'],
2847 description
=> "Maximum memory in bytes.",
2850 renderer
=> 'bytes',
2853 description
=> "Root disk size in bytes.",
2856 renderer
=> 'bytes',
2859 description
=> "VM name.",
2864 description
=> "VM run state from the 'query-status' QMP monitor command.",
2869 description
=> "PID of running qemu process.",
2874 description
=> "Uptime.",
2877 renderer
=> 'duration',
2880 description
=> "Maximum usable CPUs.",
2885 description
=> "The current config lock, if any.",
2890 description
=> "The current configured tags, if any",
2894 'running-machine' => {
2895 description
=> "The currently running machine type (if running).",
2900 description
=> "The currently running QEMU version (if running).",
2906 my $last_proc_pid_stat;
2908 # get VM status information
2909 # This must be fast and should not block ($full == false)
2910 # We only query KVM using QMP if $full == true (this can be slow)
2912 my ($opt_vmid, $full) = @_;
2916 my $storecfg = PVE
::Storage
::config
();
2918 my $list = vzlist
();
2919 my $defaults = load_defaults
();
2921 my ($uptime) = PVE
::ProcFSTools
::read_proc_uptime
(1);
2923 my $cpucount = $cpuinfo->{cpus
} || 1;
2925 foreach my $vmid (keys %$list) {
2926 next if $opt_vmid && ($vmid ne $opt_vmid);
2928 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2930 my $d = { vmid
=> int($vmid) };
2931 $d->{pid
} = int($list->{$vmid}->{pid
}) if $list->{$vmid}->{pid
};
2933 # fixme: better status?
2934 $d->{status
} = $list->{$vmid}->{pid
} ?
'running' : 'stopped';
2936 my $size = PVE
::QemuServer
::Drive
::bootdisk_size
($storecfg, $conf);
2937 if (defined($size)) {
2938 $d->{disk
} = 0; # no info available
2939 $d->{maxdisk
} = $size;
2945 $d->{cpus
} = ($conf->{sockets
} || $defaults->{sockets
})
2946 * ($conf->{cores
} || $defaults->{cores
});
2947 $d->{cpus
} = $cpucount if $d->{cpus
} > $cpucount;
2948 $d->{cpus
} = $conf->{vcpus
} if $conf->{vcpus
};
2950 $d->{name
} = $conf->{name
} || "VM $vmid";
2951 $d->{maxmem
} = $conf->{memory
} ?
$conf->{memory
}*(1024*1024)
2952 : $defaults->{memory
}*(1024*1024);
2954 if ($conf->{balloon
}) {
2955 $d->{balloon_min
} = $conf->{balloon
}*(1024*1024);
2956 $d->{shares
} = defined($conf->{shares
}) ?
$conf->{shares
}
2957 : $defaults->{shares
};
2968 $d->{diskwrite
} = 0;
2970 $d->{template
} = 1 if PVE
::QemuConfig-
>is_template($conf);
2972 $d->{serial
} = 1 if conf_has_serial
($conf);
2973 $d->{lock} = $conf->{lock} if $conf->{lock};
2974 $d->{tags
} = $conf->{tags
} if defined($conf->{tags
});
2979 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
2980 foreach my $dev (keys %$netdev) {
2981 next if $dev !~ m/^tap([1-9]\d*)i/;
2983 my $d = $res->{$vmid};
2986 $d->{netout
} += $netdev->{$dev}->{receive
};
2987 $d->{netin
} += $netdev->{$dev}->{transmit
};
2990 $d->{nics
}->{$dev}->{netout
} = int($netdev->{$dev}->{receive
});
2991 $d->{nics
}->{$dev}->{netin
} = int($netdev->{$dev}->{transmit
});
2996 my $ctime = gettimeofday
;
2998 foreach my $vmid (keys %$list) {
3000 my $d = $res->{$vmid};
3001 my $pid = $d->{pid
};
3004 my $pstat = PVE
::ProcFSTools
::read_proc_pid_stat
($pid);
3005 next if !$pstat; # not running
3007 my $used = $pstat->{utime} + $pstat->{stime
};
3009 $d->{uptime
} = int(($uptime - $pstat->{starttime
})/$cpuinfo->{user_hz
});
3011 if ($pstat->{vsize
}) {
3012 $d->{mem
} = int(($pstat->{rss
}/$pstat->{vsize
})*$d->{maxmem
});
3015 my $old = $last_proc_pid_stat->{$pid};
3017 $last_proc_pid_stat->{$pid} = {
3025 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
3027 if ($dtime > 1000) {
3028 my $dutime = $used - $old->{used
};
3030 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
3031 $last_proc_pid_stat->{$pid} = {
3037 $d->{cpu
} = $old->{cpu
};
3041 return $res if !$full;
3043 my $qmpclient = PVE
::QMPClient-
>new();
3045 my $ballooncb = sub {
3046 my ($vmid, $resp) = @_;
3048 my $info = $resp->{'return'};
3049 return if !$info->{max_mem
};
3051 my $d = $res->{$vmid};
3053 # use memory assigned to VM
3054 $d->{maxmem
} = $info->{max_mem
};
3055 $d->{balloon
} = $info->{actual
};
3057 if (defined($info->{total_mem
}) && defined($info->{free_mem
})) {
3058 $d->{mem
} = $info->{total_mem
} - $info->{free_mem
};
3059 $d->{freemem
} = $info->{free_mem
};
3062 $d->{ballooninfo
} = $info;
3065 my $blockstatscb = sub {
3066 my ($vmid, $resp) = @_;
3067 my $data = $resp->{'return'} || [];
3068 my $totalrdbytes = 0;
3069 my $totalwrbytes = 0;
3071 for my $blockstat (@$data) {
3072 $totalrdbytes = $totalrdbytes + $blockstat->{stats
}->{rd_bytes
};
3073 $totalwrbytes = $totalwrbytes + $blockstat->{stats
}->{wr_bytes
};
3075 $blockstat->{device
} =~ s/drive-//;
3076 $res->{$vmid}->{blockstat
}->{$blockstat->{device
}} = $blockstat->{stats
};
3078 $res->{$vmid}->{diskread
} = $totalrdbytes;
3079 $res->{$vmid}->{diskwrite
} = $totalwrbytes;
3082 my $machinecb = sub {
3083 my ($vmid, $resp) = @_;
3084 my $data = $resp->{'return'} || [];
3086 $res->{$vmid}->{'running-machine'} =
3087 PVE
::QemuServer
::Machine
::current_from_query_machines
($data);
3090 my $versioncb = sub {
3091 my ($vmid, $resp) = @_;
3092 my $data = $resp->{'return'} // {};
3093 my $version = 'unknown';
3095 if (my $v = $data->{qemu
}) {
3096 $version = $v->{major
} . "." . $v->{minor
} . "." . $v->{micro
};
3099 $res->{$vmid}->{'running-qemu'} = $version;
3102 my $statuscb = sub {
3103 my ($vmid, $resp) = @_;
3105 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3106 $qmpclient->queue_cmd($vmid, $machinecb, 'query-machines');
3107 $qmpclient->queue_cmd($vmid, $versioncb, 'query-version');
3108 # this fails if ballon driver is not loaded, so this must be
3109 # the last commnand (following command are aborted if this fails).
3110 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3112 my $status = 'unknown';
3113 if (!defined($status = $resp->{'return'}->{status
})) {
3114 warn "unable to get VM status\n";
3118 $res->{$vmid}->{qmpstatus
} = $resp->{'return'}->{status
};
3121 foreach my $vmid (keys %$list) {
3122 next if $opt_vmid && ($vmid ne $opt_vmid);
3123 next if !$res->{$vmid}->{pid
}; # not running
3124 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3127 $qmpclient->queue_execute(undef, 2);
3129 foreach my $vmid (keys %$list) {
3130 next if $opt_vmid && ($vmid ne $opt_vmid);
3131 next if !$res->{$vmid}->{pid
}; #not running
3133 # we can't use the $qmpclient since it might have already aborted on
3134 # 'query-balloon', but this might also fail for older versions...
3135 my $qemu_support = eval { mon_cmd
($vmid, "query-proxmox-support") };
3136 $res->{$vmid}->{'proxmox-support'} = $qemu_support // {};
3139 foreach my $vmid (keys %$list) {
3140 next if $opt_vmid && ($vmid ne $opt_vmid);
3141 $res->{$vmid}->{qmpstatus
} = $res->{$vmid}->{status
} if !$res->{$vmid}->{qmpstatus
};
3147 sub conf_has_serial
{
3150 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3151 if ($conf->{"serial$i"}) {
3159 sub conf_has_audio
{
3160 my ($conf, $id) = @_;
3163 my $audio = $conf->{"audio$id"};
3164 return if !defined($audio);
3166 my $audioproperties = parse_property_string
($audio_fmt, $audio);
3167 my $audiodriver = $audioproperties->{driver
} // 'spice';
3170 dev
=> $audioproperties->{device
},
3171 dev_id
=> "audiodev$id",
3172 backend
=> $audiodriver,
3173 backend_id
=> "$audiodriver-backend${id}",
3178 my ($audio, $audiopciaddr, $machine_version) = @_;
3182 my $id = $audio->{dev_id
};
3184 if (min_version
($machine_version, 4, 2)) {
3185 $audiodev = ",audiodev=$audio->{backend_id}";
3188 if ($audio->{dev
} eq 'AC97') {
3189 push @$devs, '-device', "AC97,id=${id}${audiopciaddr}$audiodev";
3190 } elsif ($audio->{dev
} =~ /intel\-hda$/) {
3191 push @$devs, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3192 push @$devs, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0$audiodev";
3193 push @$devs, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1$audiodev";
3195 die "unkown audio device '$audio->{dev}', implement me!";
3198 push @$devs, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3206 socket => "/var/run/qemu-server/$vmid.swtpm",
3207 pid
=> "/var/run/qemu-server/$vmid.swtpm.pid",
3211 sub add_tpm_device
{
3212 my ($vmid, $devices, $conf) = @_;
3214 return if !$conf->{tpmstate0
};
3216 my $paths = get_tpm_paths
($vmid);
3218 push @$devices, "-chardev", "socket,id=tpmchar,path=$paths->{socket}";
3219 push @$devices, "-tpmdev", "emulator,id=tpmdev,chardev=tpmchar";
3220 push @$devices, "-device", "tpm-tis,tpmdev=tpmdev";
3224 my ($storecfg, $vmid, $tpmdrive, $migration) = @_;
3226 return if !$tpmdrive;
3229 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
3230 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
3232 $state = PVE
::Storage
::map_volume
($storecfg, $tpm->{file
});
3234 $state = $tpm->{file
};
3237 my $paths = get_tpm_paths
($vmid);
3239 # during migration, we will get state from remote
3242 # run swtpm_setup to create a new TPM state if it doesn't exist yet
3249 "--create-platform-cert",
3252 "/etc/swtpm_setup.conf", # do not use XDG configs
3254 "0", # force creation as root, error if not possible
3255 "--not-overwrite", # ignore existing state, do not modify
3258 push @$setup_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3259 # TPM 2.0 supports ECC crypto, use if possible
3260 push @$setup_cmd, "--ecc" if $tpm->{version
} eq 'v2.0';
3262 run_command
($setup_cmd, outfunc
=> sub {
3263 print "swtpm_setup: $1\n";
3267 # Used to distinguish different invocations in the log.
3268 my $log_prefix = "[id=" . int(time()) . "] ";
3270 my $emulator_cmd = [
3274 "backend-uri=file://$state,mode=0600",
3276 "type=unixio,path=$paths->{socket},mode=0600",
3278 "file=$paths->{pid}",
3279 "--terminate", # terminate on QEMU disconnect
3282 "file=/run/qemu-server/$vmid-swtpm.log,level=1,prefix=$log_prefix",
3284 push @$emulator_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3285 run_command
($emulator_cmd, outfunc
=> sub { print $1; });
3287 my $tries = 100; # swtpm may take a bit to start before daemonizing, wait up to 5s for pid
3288 while (! -e
$paths->{pid
}) {
3289 die "failed to start swtpm: pid file '$paths->{pid}' wasn't created.\n" if --$tries == 0;
3293 # return untainted PID of swtpm daemon so it can be killed on error
3294 file_read_firstline
($paths->{pid
}) =~ m/(\d+)/;
3298 sub vga_conf_has_spice
{
3301 my $vgaconf = parse_vga
($vga);
3302 my $vgatype = $vgaconf->{type
};
3303 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3310 return get_host_arch
() eq $arch;
3315 return $conf->{arch
} // get_host_arch
();
3318 my $default_machines = {
3323 sub get_installed_machine_version
{
3324 my ($kvmversion) = @_;
3325 $kvmversion = kvm_user_version
() if !defined($kvmversion);
3326 $kvmversion =~ m/^(\d+\.\d+)/;
3330 sub windows_get_pinned_machine_version
{
3331 my ($machine, $base_version, $kvmversion) = @_;
3333 my $pin_version = $base_version;
3334 if (!defined($base_version) ||
3335 !PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($base_version, $kvmversion)
3337 $pin_version = get_installed_machine_version
($kvmversion);
3339 if (!$machine || $machine eq 'pc') {
3340 $machine = "pc-i440fx-$pin_version";
3341 } elsif ($machine eq 'q35') {
3342 $machine = "pc-q35-$pin_version";
3343 } elsif ($machine eq 'virt') {
3344 $machine = "virt-$pin_version";
3346 warn "unknown machine type '$machine', not touching that!\n";
3352 sub get_vm_machine
{
3353 my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
3355 my $machine = $forcemachine || $conf->{machine
};
3357 if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
3358 $kvmversion //= kvm_user_version
();
3359 # we must pin Windows VMs without a specific version to 5.1, as 5.2 fixed a bug in ACPI
3360 # layout which confuses windows quite a bit and may result in various regressions..
3361 # see: https://lists.gnu.org/archive/html/qemu-devel/2021-02/msg08484.html
3362 if (windows_version
($conf->{ostype
})) {
3363 $machine = windows_get_pinned_machine_version
($machine, '5.1', $kvmversion);
3366 $machine ||= $default_machines->{$arch};
3367 if ($add_pve_version) {
3368 my $pvever = PVE
::QemuServer
::Machine
::get_pve_version
($kvmversion);
3369 $machine .= "+pve$pvever";
3373 if ($add_pve_version && $machine !~ m/\+pve\d+?(?:\.pxe)?$/) {
3374 my $is_pxe = $machine =~ m/^(.*?)\.pxe$/;
3375 $machine = $1 if $is_pxe;
3377 # for version-pinned machines that do not include a pve-version (e.g.
3378 # pc-q35-4.1), we assume 0 to keep them stable in case we bump
3379 $machine .= '+pve0';
3381 $machine .= '.pxe' if $is_pxe;
3387 sub get_ovmf_files
($$$) {
3388 my ($arch, $efidisk, $smm) = @_;
3390 my $types = $OVMF->{$arch}
3391 or die "no OVMF images known for architecture '$arch'\n";
3393 my $type = 'default';
3394 if ($arch ne "aarch64" && defined($efidisk->{efitype
}) && $efidisk->{efitype
} eq '4m') {
3395 $type = $smm ?
"4m" : "4m-no-smm";
3396 $type .= '-ms' if $efidisk->{'pre-enrolled-keys'};
3399 my ($ovmf_code, $ovmf_vars) = $types->{$type}->@*;
3400 die "EFI base image '$ovmf_code' not found\n" if ! -f
$ovmf_code;
3401 die "EFI vars image '$ovmf_vars' not found\n" if ! -f
$ovmf_vars;
3403 return ($ovmf_code, $ovmf_vars);
3407 aarch64
=> '/usr/bin/qemu-system-aarch64',
3408 x86_64
=> '/usr/bin/qemu-system-x86_64',
3410 sub get_command_for_arch
($) {
3412 return '/usr/bin/kvm' if is_native
($arch);
3414 my $cmd = $Arch2Qemu->{$arch}
3415 or die "don't know how to emulate architecture '$arch'\n";
3419 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
3420 # to use in a QEMU command line (-cpu element), first array_intersect the result
3421 # of query_supported_ with query_understood_. This is necessary because:
3423 # a) query_understood_ returns flags the host cannot use and
3424 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
3425 # flags, but CPU settings - with most of them being flags. Those settings
3426 # (and some flags, curiously) cannot be specified as a "-cpu" argument.
3428 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
3429 # expensive. If you need the value returned from this, you can get it much
3430 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
3431 # $accel being 'kvm' or 'tcg'.
3433 # pvestatd calls this function on startup and whenever the QEMU/KVM version
3434 # changes, automatically populating pmxcfs.
3436 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
3437 # since kvm and tcg machines support different flags
3439 sub query_supported_cpu_flags
{
3442 $arch //= get_host_arch
();
3443 my $default_machine = $default_machines->{$arch};
3447 # FIXME: Once this is merged, the code below should work for ARM as well:
3448 # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
3449 die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
3452 my $kvm_supported = defined(kvm_version
());
3453 my $qemu_cmd = get_command_for_arch
($arch);
3455 my $pidfile = PVE
::QemuServer
::Helpers
::pidfile_name
($fakevmid);
3457 # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
3458 my $query_supported_run_qemu = sub {
3464 '-machine', $default_machine,
3466 '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server=on,wait=off",
3467 '-mon', 'chardev=qmp,mode=control',
3468 '-pidfile', $pidfile,
3473 push @$cmd, '-accel', 'tcg';
3476 my $rc = run_command
($cmd, noerr
=> 1, quiet
=> 0);
3477 die "QEMU flag querying VM exited with code " . $rc if $rc;
3480 my $cmd_result = mon_cmd
(
3482 'query-cpu-model-expansion',
3484 model
=> { name
=> 'host' }
3487 my $props = $cmd_result->{model
}->{props
};
3488 foreach my $prop (keys %$props) {
3489 next if $props->{$prop} ne '1';
3490 # QEMU returns some flags multiple times, with '_', '.' or '-'
3491 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
3492 # We only keep those with underscores, to match /proc/cpuinfo
3493 $prop =~ s/\.|-/_/g;
3494 $flags->{$prop} = 1;
3499 # force stop with 10 sec timeout and 'nocheck', always stop, even if QMP failed
3500 vm_stop
(undef, $fakevmid, 1, 1, 10, 0, 1);
3504 return [ sort keys %$flags ];
3507 # We need to query QEMU twice, since KVM and TCG have different supported flags
3508 PVE
::QemuConfig-
>lock_config($fakevmid, sub {
3509 $flags->{tcg
} = eval { $query_supported_run_qemu->(0) };
3510 warn "warning: failed querying supported tcg flags: $@\n" if $@;
3512 if ($kvm_supported) {
3513 $flags->{kvm
} = eval { $query_supported_run_qemu->(1) };
3514 warn "warning: failed querying supported kvm flags: $@\n" if $@;
3521 # Understood CPU flags are written to a file at 'pve-qemu' compile time
3522 my $understood_cpu_flag_dir = "/usr/share/kvm";
3523 sub query_understood_cpu_flags
{
3524 my $arch = get_host_arch
();
3525 my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
3527 die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
3530 my $raw = file_get_contents
($filepath);
3531 $raw =~ s/^\s+|\s+$//g;
3532 my @flags = split(/\s+/, $raw);
3537 # Since commit 277d33454f77ec1d1e0bc04e37621e4dd2424b67 in pve-qemu, smm is not off by default
3538 # anymore. But smm=off seems to be required when using SeaBIOS and serial display.
3539 my sub should_disable_smm
{
3540 my ($conf, $vga, $machine) = @_;
3542 return if $machine =~ m/^virt/; # there is no smm flag that could be disabled
3544 return (!defined($conf->{bios
}) || $conf->{bios
} eq 'seabios') &&
3545 $vga->{type
} && $vga->{type
} =~ m/^(serial\d+|none)$/;
3548 my sub print_ovmf_drive_commandlines
{
3549 my ($conf, $storecfg, $vmid, $arch, $q35, $version_guard) = @_;
3551 my $d = $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
3553 my ($ovmf_code, $ovmf_vars) = get_ovmf_files
($arch, $d, $q35);
3555 my $var_drive_str = "if=pflash,unit=1,id=drive-efidisk0";
3557 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($d->{file
}, 1);
3558 my ($path, $format) = $d->@{'file', 'format'};
3560 $path = PVE
::Storage
::path
($storecfg, $d->{file
});
3561 if (!defined($format)) {
3562 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
3563 $format = qemu_img_format
($scfg, $volname);
3565 } elsif (!defined($format)) {
3566 die "efidisk format must be specified\n";
3568 # SPI flash does lots of read-modify-write OPs, without writeback this gets really slow #3329
3569 if ($path =~ m/^rbd:/) {
3570 $var_drive_str .= ',cache=writeback';
3571 $path .= ':rbd_cache_policy=writeback'; # avoid write-around, we *need* to cache writes too
3573 $var_drive_str .= ",format=$format,file=$path";
3575 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $format eq 'raw' && $version_guard->(4, 1, 2);
3576 $var_drive_str .= ',readonly=on' if drive_is_read_only
($conf, $d);
3578 log_warn
("no efidisk configured! Using temporary efivars disk.");
3579 my $path = "/tmp/$vmid-ovmf.fd";
3580 PVE
::Tools
::file_copy
($ovmf_vars, $path, -s
$ovmf_vars);
3581 $var_drive_str .= ",format=raw,file=$path";
3582 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $version_guard->(4, 1, 2);
3585 return ("if=pflash,unit=0,format=raw,readonly=on,file=$ovmf_code", $var_drive_str);
3588 sub config_to_command
{
3589 my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu,
3592 my ($globalFlags, $machineFlags, $rtcFlags) = ([], [], []);
3595 my $ostype = $conf->{ostype
};
3596 my $winversion = windows_version
($ostype);
3597 my $kvm = $conf->{kvm
};
3598 my $nodename = nodename
();
3600 my $arch = get_vm_arch
($conf);
3601 my $kvm_binary = get_command_for_arch
($arch);
3602 my $kvmver = kvm_user_version
($kvm_binary);
3604 if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
3605 $kvmver //= "undefined";
3606 die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
3609 my $add_pve_version = min_version
($kvmver, 4, 1);
3611 my $machine_type = get_vm_machine
($conf, $forcemachine, $arch, $add_pve_version);
3612 my $machine_version = extract_version
($machine_type, $kvmver);
3613 $kvm //= 1 if is_native
($arch);
3615 $machine_version =~ m/(\d+)\.(\d+)/;
3616 my ($machine_major, $machine_minor) = ($1, $2);
3618 if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
3619 warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
3620 } elsif (!min_version
($kvmver, $machine_major, $machine_minor)) {
3621 die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type',"
3622 ." please upgrade node '$nodename'\n"
3623 } elsif (!PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($machine_version, $kvmver)) {
3624 my $max_pve_version = PVE
::QemuServer
::Machine
::get_pve_version
($machine_version);
3625 die "Installed qemu-server (max feature level for $machine_major.$machine_minor is"
3626 ." pve$max_pve_version) is too old to run machine type '$machine_type', please upgrade"
3627 ." node '$nodename'\n";
3630 # if a specific +pve version is required for a feature, use $version_guard
3631 # instead of min_version to allow machines to be run with the minimum
3633 my $required_pve_version = 0;
3634 my $version_guard = sub {
3635 my ($major, $minor, $pve) = @_;
3636 return 0 if !min_version
($machine_version, $major, $minor, $pve);
3637 my $max_pve = PVE
::QemuServer
::Machine
::get_pve_version
("$major.$minor");
3638 return 1 if min_version
($machine_version, $major, $minor, $max_pve+1);
3639 $required_pve_version = $pve if $pve && $pve > $required_pve_version;
3643 if ($kvm && !defined kvm_version
()) {
3644 die "KVM virtualisation configured, but not available. Either disable in VM configuration"
3645 ." or enable in BIOS.\n";
3648 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
3649 my $hotplug_features = parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
3650 my $use_old_bios_files = undef;
3651 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
3654 if ($conf->{affinity
}) {
3655 push @$cmd, '/usr/bin/taskset', '--cpu-list', '--all-tasks', $conf->{affinity
};
3658 push @$cmd, $kvm_binary;
3660 push @$cmd, '-id', $vmid;
3662 my $vmname = $conf->{name
} || "vm$vmid";
3664 push @$cmd, '-name', "$vmname,debug-threads=on";
3666 push @$cmd, '-no-shutdown';
3670 my $qmpsocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid);
3671 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server=on,wait=off";
3672 push @$cmd, '-mon', "chardev=qmp,mode=control";
3674 if (min_version
($machine_version, 2, 12)) {
3675 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3676 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3679 push @$cmd, '-pidfile' , PVE
::QemuServer
::Helpers
::pidfile_name
($vmid);
3681 push @$cmd, '-daemonize';
3683 if ($conf->{smbios1
}) {
3684 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
3685 if ($smbios_conf->{base64
}) {
3686 # Do not pass base64 flag to qemu
3687 delete $smbios_conf->{base64
};
3688 my $smbios_string = "";
3689 foreach my $key (keys %$smbios_conf) {
3691 if ($key eq "uuid") {
3692 $value = $smbios_conf->{uuid
}
3694 $value = decode_base64
($smbios_conf->{$key});
3696 # qemu accepts any binary data, only commas need escaping by double comma
3698 $smbios_string .= "," . $key . "=" . $value if $value;
3700 push @$cmd, '-smbios', "type=1" . $smbios_string;
3702 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3706 if ($conf->{bios
} && $conf->{bios
} eq 'ovmf') {
3707 my ($code_drive_str, $var_drive_str) =
3708 print_ovmf_drive_commandlines
($conf, $storecfg, $vmid, $arch, $q35, $version_guard);
3709 push $cmd->@*, '-drive', $code_drive_str;
3710 push $cmd->@*, '-drive', $var_drive_str;
3713 if ($q35) { # tell QEMU to load q35 config early
3714 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3715 if (min_version
($machine_version, 4, 0)) {
3716 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3718 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3722 if (defined(my $fixups = qemu_created_version_fixups
($conf, $forcemachine, $kvmver))) {
3723 push @$cmd, $fixups->@*;
3726 if ($conf->{vmgenid
}) {
3727 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid
};
3730 # add usb controllers
3731 my @usbcontrollers = PVE
::QemuServer
::USB
::get_usb_controllers
(
3732 $conf, $bridges, $arch, $machine_type, $machine_version);
3733 push @$devices, @usbcontrollers if @usbcontrollers;
3734 my $vga = parse_vga
($conf->{vga
});
3736 my $qxlnum = vga_conf_has_spice
($conf->{vga
});
3737 $vga->{type
} = 'qxl' if $qxlnum;
3739 if (!$vga->{type
}) {
3740 if ($arch eq 'aarch64') {
3741 $vga->{type
} = 'virtio';
3742 } elsif (min_version
($machine_version, 2, 9)) {
3743 $vga->{type
} = (!$winversion || $winversion >= 6) ?
'std' : 'cirrus';
3745 $vga->{type
} = ($winversion >= 6) ?
'std' : 'cirrus';
3749 # enable absolute mouse coordinates (needed by vnc)
3750 my $tablet = $conf->{tablet
};
3751 if (!defined($tablet)) {
3752 $tablet = $defaults->{tablet
};
3753 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3754 $tablet = 0 if $vga->{type
} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3758 push @$devices, '-device', print_tabletdevice_full
($conf, $arch) if $tablet;
3759 my $kbd = print_keyboarddevice_full
($conf, $arch);
3760 push @$devices, '-device', $kbd if defined($kbd);
3763 my $bootorder = device_bootorder
($conf);
3765 # host pci device passthrough
3766 my ($kvm_off, $gpu_passthrough, $legacy_igd, $pci_devices) = PVE
::QemuServer
::PCI
::print_hostpci_devices
(
3767 $vmid, $conf, $devices, $vga, $winversion, $bridges, $arch, $machine_type, $bootorder);
3770 my $usb_dev_features = {};
3771 $usb_dev_features->{spice_usb3
} = 1 if min_version
($machine_version, 4, 0);
3773 my @usbdevices = PVE
::QemuServer
::USB
::get_usb_devices
(
3774 $conf, $usb_dev_features, $bootorder, $machine_version);
3775 push @$devices, @usbdevices if @usbdevices;
3778 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3779 my $path = $conf->{"serial$i"} or next;
3780 if ($path eq 'socket') {
3781 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3782 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server=on,wait=off";
3783 # On aarch64, serial0 is the UART device. QEMU only allows
3784 # connecting UART devices via the '-serial' command line, as
3785 # the device has a fixed slot on the hardware...
3786 if ($arch eq 'aarch64' && $i == 0) {
3787 push @$devices, '-serial', "chardev:serial$i";
3789 push @$devices, '-device', "isa-serial,chardev=serial$i";
3792 die "no such serial device\n" if ! -c
$path;
3793 push @$devices, '-chardev', "serial,id=serial$i,path=$path";
3794 push @$devices, '-device', "isa-serial,chardev=serial$i";
3799 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3800 if (my $path = $conf->{"parallel$i"}) {
3801 die "no such parallel device\n" if ! -c
$path;
3802 my $devtype = $path =~ m!^/dev/usb/lp! ?
'serial' : 'parallel';
3803 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3804 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3808 if (min_version
($machine_version, 4, 0) && (my $audio = conf_has_audio
($conf))) {
3809 my $audiopciaddr = print_pci_addr
("audio0", $bridges, $arch, $machine_type);
3810 my $audio_devs = audio_devs
($audio, $audiopciaddr, $machine_version);
3811 push @$devices, @$audio_devs;
3814 add_tpm_device
($vmid, $devices, $conf);
3817 $sockets = $conf->{smp
} if $conf->{smp
}; # old style - no longer iused
3818 $sockets = $conf->{sockets
} if $conf->{sockets
};
3820 my $cores = $conf->{cores
} || 1;
3822 my $maxcpus = $sockets * $cores;
3824 my $vcpus = $conf->{vcpus
} ?
$conf->{vcpus
} : $maxcpus;
3826 my $allowed_vcpus = $cpuinfo->{cpus
};
3828 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n" if ($allowed_vcpus < $maxcpus);
3830 if ($hotplug_features->{cpu
} && min_version
($machine_version, 2, 7)) {
3831 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3832 for (my $i = 2; $i <= $vcpus; $i++) {
3833 my $cpustr = print_cpu_device
($conf,$i);
3834 push @$cmd, '-device', $cpustr;
3839 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3841 push @$cmd, '-nodefaults';
3843 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3845 push $machineFlags->@*, 'acpi=off' if defined($conf->{acpi
}) && $conf->{acpi
} == 0;
3847 push @$cmd, '-no-reboot' if defined($conf->{reboot
}) && $conf->{reboot
} == 0;
3849 if ($vga->{type
} && $vga->{type
} !~ m/^serial\d+$/ && $vga->{type
} ne 'none'){
3850 push @$devices, '-device', print_vga_device
(
3851 $conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3853 push @$cmd, '-display', 'egl-headless,gl=core' if $vga->{type
} eq 'virtio-gl'; # VIRGL
3855 my $socket = PVE
::QemuServer
::Helpers
::vnc_socket
($vmid);
3856 push @$cmd, '-vnc', "unix:$socket,password=on";
3858 push @$cmd, '-vga', 'none' if $vga->{type
} eq 'none';
3859 push @$cmd, '-nographic';
3863 my $tdf = defined($conf->{tdf
}) ?
$conf->{tdf
} : $defaults->{tdf
};
3864 my $useLocaltime = $conf->{localtime};
3866 if ($winversion >= 5) { # windows
3867 $useLocaltime = 1 if !defined($conf->{localtime});
3869 # use time drift fix when acpi is enabled
3870 if (!(defined($conf->{acpi
}) && $conf->{acpi
} == 0)) {
3871 $tdf = 1 if !defined($conf->{tdf
});
3875 if ($winversion >= 6) {
3876 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3877 push @$machineFlags, 'hpet=off';
3880 push @$rtcFlags, 'driftfix=slew' if $tdf;
3882 if ($conf->{startdate
} && $conf->{startdate
} ne 'now') {
3883 push @$rtcFlags, "base=$conf->{startdate}";
3884 } elsif ($useLocaltime) {
3885 push @$rtcFlags, 'base=localtime';
3889 push @$cmd, '-cpu', $forcecpu;
3891 push @$cmd, get_cpu_options
($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough);
3894 PVE
::QemuServer
::Memory
::config
(
3895 $conf, $vmid, $sockets, $cores, $defaults, $hotplug_features->{memory
}, $cmd);
3897 push @$cmd, '-S' if $conf->{freeze
};
3899 push @$cmd, '-k', $conf->{keyboard
} if defined($conf->{keyboard
});
3901 my $guest_agent = parse_guest_agent
($conf);
3903 if ($guest_agent->{enabled
}) {
3904 my $qgasocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid, 1);
3905 push @$devices, '-chardev', "socket,path=$qgasocket,server=on,wait=off,id=qga0";
3907 if (!$guest_agent->{type
} || $guest_agent->{type
} eq 'virtio') {
3908 my $pciaddr = print_pci_addr
("qga0", $bridges, $arch, $machine_type);
3909 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3910 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3911 } elsif ($guest_agent->{type
} eq 'isa') {
3912 push @$devices, '-device', "isa-serial,chardev=qga0";
3916 my $rng = $conf->{rng0
} ? parse_rng
($conf->{rng0
}) : undef;
3917 if ($rng && $version_guard->(4, 1, 2)) {
3918 check_rng_source
($rng->{source
});
3920 my $max_bytes = $rng->{max_bytes
} // $rng_fmt->{max_bytes
}->{default};
3921 my $period = $rng->{period
} // $rng_fmt->{period
}->{default};
3922 my $limiter_str = "";
3924 $limiter_str = ",max-bytes=$max_bytes,period=$period";
3927 my $rng_addr = print_pci_addr
("rng0", $bridges, $arch, $machine_type);
3928 push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3929 push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3934 if ($qxlnum || $vga->{type
} =~ /^virtio/) {
3937 for (my $i = 1; $i < $qxlnum; $i++){
3938 push @$devices, '-device', print_vga_device
(
3939 $conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3942 # assume other OS works like Linux
3943 my ($ram, $vram) = ("134217728", "67108864");
3944 if ($vga->{memory
}) {
3945 $ram = PVE
::Tools
::convert_size
($qxlnum*4*$vga->{memory
}, 'mb' => 'b');
3946 $vram = PVE
::Tools
::convert_size
($qxlnum*2*$vga->{memory
}, 'mb' => 'b');
3948 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3949 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3953 my $pciaddr = print_pci_addr
("spice", $bridges, $arch, $machine_type);
3955 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
3956 my @nodeaddrs = PVE
::Tools
::getaddrinfo_all
('localhost', family
=> $pfamily);
3957 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3959 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3960 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3961 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3963 my $localhost = PVE
::Network
::addr_to_ip
($nodeaddrs[0]->{addr
});
3964 $spice_port = PVE
::Tools
::next_spice_port
($pfamily, $localhost);
3966 my $spice_enhancement_str = $conf->{spice_enhancements
} // '';
3967 my $spice_enhancement = parse_property_string
($spice_enhancements_fmt, $spice_enhancement_str);
3968 if ($spice_enhancement->{foldersharing
}) {
3969 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3970 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3973 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3974 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}"
3975 if $spice_enhancement->{videostreaming
};
3977 push @$devices, '-spice', "$spice_opts";
3980 # enable balloon by default, unless explicitly disabled
3981 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
3982 my $pciaddr = print_pci_addr
("balloon0", $bridges, $arch, $machine_type);
3983 my $ballooncmd = "virtio-balloon-pci,id=balloon0$pciaddr";
3984 $ballooncmd .= ",free-page-reporting=on" if min_version
($machine_version, 6, 2);
3985 push @$devices, '-device', $ballooncmd;
3988 if ($conf->{watchdog
}) {
3989 my $wdopts = parse_watchdog
($conf->{watchdog
});
3990 my $pciaddr = print_pci_addr
("watchdog", $bridges, $arch, $machine_type);
3991 my $watchdog = $wdopts->{model
} || 'i6300esb';
3992 push @$devices, '-device', "$watchdog$pciaddr";
3993 push @$devices, '-watchdog-action', $wdopts->{action
} if $wdopts->{action
};
3997 my $scsicontroller = {};
3998 my $ahcicontroller = {};
3999 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : $defaults->{scsihw
};
4001 # Add iscsi initiator name if available
4002 if (my $initiator = get_initiator_name
()) {
4003 push @$devices, '-iscsi', "initiator-name=$initiator";
4006 PVE
::QemuConfig-
>foreach_volume($conf, sub {
4007 my ($ds, $drive) = @_;
4009 if (PVE
::Storage
::parse_volume_id
($drive->{file
}, 1)) {
4010 check_volume_storage_type
($storecfg, $drive->{file
});
4011 push @$vollist, $drive->{file
};
4014 # ignore efidisk here, already added in bios/fw handling code above
4015 return if $drive->{interface
} eq 'efidisk';
4017 return if $drive->{interface
} eq 'tpmstate';
4019 $use_virtio = 1 if $ds =~ m/^virtio/;
4021 $drive->{bootindex
} = $bootorder->{$ds} if $bootorder->{$ds};
4023 if ($drive->{interface
} eq 'virtio'){
4024 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread
};
4027 if ($drive->{interface
} eq 'scsi') {
4029 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
4031 die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
4032 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
4034 my $pciaddr = print_pci_addr
("$controller_prefix$controller", $bridges, $arch, $machine_type);
4035 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ?
"virtio-scsi-pci" : $scsihw;
4038 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{iothread
}){
4039 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4040 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4041 } elsif ($drive->{iothread
}) {
4043 "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n"
4048 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{queues
}){
4049 $queues = ",num_queues=$drive->{queues}";
4052 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues"
4053 if !$scsicontroller->{$controller};
4054 $scsicontroller->{$controller}=1;
4057 if ($drive->{interface
} eq 'sata') {
4058 my $controller = int($drive->{index} / $PVE::QemuServer
::Drive
::MAX_SATA_DISKS
);
4059 my $pciaddr = print_pci_addr
("ahci$controller", $bridges, $arch, $machine_type);
4060 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr"
4061 if !$ahcicontroller->{$controller};
4062 $ahcicontroller->{$controller}=1;
4065 my $pbs_conf = $pbs_backing->{$ds};
4066 my $pbs_name = undef;
4068 $pbs_name = "drive-$ds-pbs";
4069 push @$devices, '-blockdev', print_pbs_blockdev
($pbs_conf, $pbs_name);
4072 my $drive_cmd = print_drive_commandline_full
(
4073 $storecfg, $vmid, $drive, $pbs_name, min_version
($kvmver, 6, 0));
4075 # extra protection for templates, but SATA and IDE don't support it..
4076 $drive_cmd .= ',readonly=on' if drive_is_read_only
($conf, $drive);
4078 push @$devices, '-drive',$drive_cmd;
4079 push @$devices, '-device', print_drivedevice_full
(
4080 $storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4083 for (my $i = 0; $i < $MAX_NETS; $i++) {
4084 my $netname = "net$i";
4086 next if !$conf->{$netname};
4087 my $d = parse_net
($conf->{$netname});
4089 # save the MAC addr here (could be auto-gen. in some odd setups) for FDB registering later?
4091 $use_virtio = 1 if $d->{model
} eq 'virtio';
4093 $d->{bootindex
} = $bootorder->{$netname} if $bootorder->{$netname};
4095 my $netdevfull = print_netdev_full
($vmid, $conf, $arch, $d, $netname);
4096 push @$devices, '-netdev', $netdevfull;
4098 my $netdevicefull = print_netdevice_full
(
4099 $vmid, $conf, $d, $netname, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version);
4101 push @$devices, '-device', $netdevicefull;
4104 if ($conf->{ivshmem
}) {
4105 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
4109 $bus = print_pcie_addr
("ivshmem");
4111 $bus = print_pci_addr
("ivshmem", $bridges, $arch, $machine_type);
4114 my $ivshmem_name = $ivshmem->{name
} // $vmid;
4115 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4117 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4118 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path"
4119 .",size=$ivshmem->{size}M";
4122 # pci.4 is nested in pci.1
4123 $bridges->{1} = 1 if $bridges->{4};
4125 if (!$q35) { # add pci bridges
4126 if (min_version
($machine_version, 2, 3)) {
4130 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4133 for my $k (sort {$b cmp $a} keys %$bridges) {
4134 next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
4137 if ($k == 2 && $legacy_igd) {
4140 my $pciaddr = print_pci_addr
("pci.$k_name", undef, $arch, $machine_type);
4141 my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
4143 if ($q35) { # add after -readconfig pve-q35.cfg
4144 splice @$devices, 2, 0, '-device', $devstr;
4146 unshift @$devices, '-device', $devstr if $k > 0;
4151 push @$machineFlags, 'accel=tcg';
4154 push @$machineFlags, 'smm=off' if should_disable_smm
($conf, $vga, $machine_type);
4156 my $machine_type_min = $machine_type;
4157 if ($add_pve_version) {
4158 $machine_type_min =~ s/\+pve\d+$//;
4159 $machine_type_min .= "+pve$required_pve_version";
4161 push @$machineFlags, "type=${machine_type_min}";
4163 push @$cmd, @$devices;
4164 push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
4165 push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
4166 push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
4168 if (my $vmstate = $conf->{vmstate
}) {
4169 my $statepath = PVE
::Storage
::path
($storecfg, $vmstate);
4170 push @$vollist, $vmstate;
4171 push @$cmd, '-loadstate', $statepath;
4172 print "activating and using '$vmstate' as vmstate\n";
4175 if (PVE
::QemuConfig-
>is_template($conf)) {
4176 # needed to workaround base volumes being read-only
4177 push @$cmd, '-snapshot';
4181 if ($conf->{args
}) {
4182 my $aa = PVE
::Tools
::split_args
($conf->{args
});
4186 return wantarray ?
($cmd, $vollist, $spice_port, $pci_devices) : $cmd;
4189 sub check_rng_source
{
4192 # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
4193 die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
4196 my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
4197 if ($source eq '/dev/hwrng' && file_read_firstline
($rng_current) eq 'none') {
4198 # Needs to abort, otherwise QEMU crashes on first rng access. Note that rng_current cannot
4199 # be changed to 'none' manually, so once the VM is past this point, it's no longer an issue.
4200 die "Cannot start VM with passed-through RNG device: '/dev/hwrng' exists, but"
4201 ." '$rng_current' is set to 'none'. Ensure that a compatible hardware-RNG is attached"
4209 my $res = mon_cmd
($vmid, 'query-spice');
4211 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4214 sub vm_devices_list
{
4217 my $res = mon_cmd
($vmid, 'query-pci');
4218 my $devices_to_check = [];
4220 foreach my $pcibus (@$res) {
4221 push @$devices_to_check, @{$pcibus->{devices
}},
4224 while (@$devices_to_check) {
4226 for my $d (@$devices_to_check) {
4227 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4228 next if !$d->{'pci_bridge'} || !$d->{'pci_bridge'}->{devices
};
4230 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices
}});
4231 push @$to_check, @{$d->{'pci_bridge'}->{devices
}};
4233 $devices_to_check = $to_check;
4236 my $resblock = mon_cmd
($vmid, 'query-block');
4237 foreach my $block (@$resblock) {
4238 if($block->{device
} =~ m/^drive-(\S+)/){
4243 my $resmice = mon_cmd
($vmid, 'query-mice');
4244 foreach my $mice (@$resmice) {
4245 if ($mice->{name
} eq 'QEMU HID Tablet') {
4246 $devices->{tablet
} = 1;
4251 # for usb devices there is no query-usb
4252 # but we can iterate over the entries in
4253 # qom-list path=/machine/peripheral
4254 my $resperipheral = mon_cmd
($vmid, 'qom-list', path
=> '/machine/peripheral');
4255 foreach my $per (@$resperipheral) {
4256 if ($per->{name
} =~ m/^usb(?:redirdev)?\d+$/) {
4257 $devices->{$per->{name
}} = 1;
4265 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4267 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
4269 my $devices_list = vm_devices_list
($vmid);
4270 return 1 if defined($devices_list->{$deviceid});
4272 # add PCI bridge if we need it for the device
4273 qemu_add_pci_bridge
($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type);
4275 if ($deviceid eq 'tablet') {
4276 qemu_deviceadd
($vmid, print_tabletdevice_full
($conf, $arch));
4277 } elsif ($deviceid eq 'keyboard') {
4278 qemu_deviceadd
($vmid, print_keyboarddevice_full
($conf, $arch));
4279 } elsif ($deviceid =~ m/^usbredirdev(\d+)$/) {
4281 qemu_spice_usbredir_chardev_add
($vmid, "usbredirchardev$id");
4282 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_spice_usbdevice
($id, "xhci", $id + 1));
4283 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4284 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_usbdevice_full
($conf, $deviceid, $device, {}, $1 + 1));
4285 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4286 qemu_iothread_add
($vmid, $deviceid, $device);
4288 qemu_driveadd
($storecfg, $vmid, $device);
4289 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4291 qemu_deviceadd
($vmid, $devicefull);
4292 eval { qemu_deviceaddverify
($vmid, $deviceid); };
4294 eval { qemu_drivedel
($vmid, $deviceid); };
4298 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4299 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : "lsi";
4300 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4301 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ?
"virtio-scsi-pci" : $scsihw;
4303 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4305 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread
}) {
4306 qemu_iothread_add
($vmid, $deviceid, $device);
4307 $devicefull .= ",iothread=iothread-$deviceid";
4310 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues
}) {
4311 $devicefull .= ",num_queues=$device->{queues}";
4314 qemu_deviceadd
($vmid, $devicefull);
4315 qemu_deviceaddverify
($vmid, $deviceid);
4316 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4317 qemu_findorcreatescsihw
($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4318 qemu_driveadd
($storecfg, $vmid, $device);
4320 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4321 eval { qemu_deviceadd
($vmid, $devicefull); };
4323 eval { qemu_drivedel
($vmid, $deviceid); };
4327 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4328 return if !qemu_netdevadd
($vmid, $conf, $arch, $device, $deviceid);
4330 my $machine_type = PVE
::QemuServer
::Machine
::qemu_machine_pxe
($vmid, $conf);
4331 my $machine_version = PVE
::QemuServer
::Machine
::extract_version
($machine_type);
4332 my $use_old_bios_files = undef;
4333 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
4335 my $netdevicefull = print_netdevice_full
(
4336 $vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type, $machine_version);
4337 qemu_deviceadd
($vmid, $netdevicefull);
4339 qemu_deviceaddverify
($vmid, $deviceid);
4340 qemu_set_link_status
($vmid, $deviceid, !$device->{link_down
});
4343 eval { qemu_netdevdel
($vmid, $deviceid); };
4347 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4349 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4350 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4352 qemu_deviceadd
($vmid, $devicefull);
4353 qemu_deviceaddverify
($vmid, $deviceid);
4355 die "can't hotplug device '$deviceid'\n";
4361 # fixme: this should raise exceptions on error!
4362 sub vm_deviceunplug
{
4363 my ($vmid, $conf, $deviceid) = @_;
4365 my $devices_list = vm_devices_list
($vmid);
4366 return 1 if !defined($devices_list->{$deviceid});
4368 my $bootdisks = PVE
::QemuServer
::Drive
::get_bootdisks
($conf);
4369 die "can't unplug bootdisk '$deviceid'\n" if grep {$_ eq $deviceid} @$bootdisks;
4371 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard' || $deviceid eq 'xhci') {
4372 qemu_devicedel
($vmid, $deviceid);
4373 } elsif ($deviceid =~ m/^usbredirdev\d+$/) {
4374 qemu_devicedel
($vmid, $deviceid);
4375 qemu_devicedelverify
($vmid, $deviceid);
4376 } elsif ($deviceid =~ m/^usb\d+$/) {
4377 qemu_devicedel
($vmid, $deviceid);
4378 qemu_devicedelverify
($vmid, $deviceid);
4379 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4380 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4382 qemu_devicedel
($vmid, $deviceid);
4383 qemu_devicedelverify
($vmid, $deviceid);
4384 qemu_drivedel
($vmid, $deviceid);
4385 qemu_iothread_del
($vmid, $deviceid, $device);
4386 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4387 qemu_devicedel
($vmid, $deviceid);
4388 qemu_devicedelverify
($vmid, $deviceid);
4389 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4390 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4392 qemu_devicedel
($vmid, $deviceid);
4393 qemu_devicedelverify
($vmid, $deviceid);
4394 qemu_drivedel
($vmid, $deviceid);
4395 qemu_deletescsihw
($conf, $vmid, $deviceid);
4397 qemu_iothread_del
($vmid, "virtioscsi$device->{index}", $device)
4398 if $conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single');
4399 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4400 qemu_devicedel
($vmid, $deviceid);
4401 qemu_devicedelverify
($vmid, $deviceid);
4402 qemu_netdevdel
($vmid, $deviceid);
4404 die "can't unplug device '$deviceid'\n";
4410 sub qemu_spice_usbredir_chardev_add
{
4411 my ($vmid, $id) = @_;
4413 mon_cmd
($vmid, "chardev-add" , (
4424 sub qemu_deviceadd
{
4425 my ($vmid, $devicefull) = @_;
4427 $devicefull = "driver=".$devicefull;
4428 my %options = split(/[=,]/, $devicefull);
4430 mon_cmd
($vmid, "device_add" , %options);
4433 sub qemu_devicedel
{
4434 my ($vmid, $deviceid) = @_;
4436 my $ret = mon_cmd
($vmid, "device_del", id
=> $deviceid);
4439 sub qemu_iothread_add
{
4440 my ($vmid, $deviceid, $device) = @_;
4442 if ($device->{iothread
}) {
4443 my $iothreads = vm_iothreads_list
($vmid);
4444 qemu_objectadd
($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4448 sub qemu_iothread_del
{
4449 my ($vmid, $deviceid, $device) = @_;
4451 if ($device->{iothread
}) {
4452 my $iothreads = vm_iothreads_list
($vmid);
4453 qemu_objectdel
($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4457 sub qemu_objectadd
{
4458 my ($vmid, $objectid, $qomtype) = @_;
4460 mon_cmd
($vmid, "object-add", id
=> $objectid, "qom-type" => $qomtype);
4465 sub qemu_objectdel
{
4466 my ($vmid, $objectid) = @_;
4468 mon_cmd
($vmid, "object-del", id
=> $objectid);
4474 my ($storecfg, $vmid, $device) = @_;
4476 my $kvmver = get_running_qemu_version
($vmid);
4477 my $io_uring = min_version
($kvmver, 6, 0);
4478 my $drive = print_drive_commandline_full
($storecfg, $vmid, $device, undef, $io_uring);
4479 $drive =~ s/\\/\\\\/g;
4480 my $ret = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "drive_add auto \"$drive\"");
4482 # If the command succeeds qemu prints: "OK
"
4483 return 1 if $ret =~ m/OK/s;
4485 die "adding drive failed
: $ret\n";
4489 my ($vmid, $deviceid) = @_;
4491 my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-
$deviceid");
4494 return 1 if $ret eq "";
4496 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4497 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4499 die "deleting drive
$deviceid failed
: $ret\n";
4502 sub qemu_deviceaddverify {
4503 my ($vmid, $deviceid) = @_;
4505 for (my $i = 0; $i <= 5; $i++) {
4506 my $devices_list = vm_devices_list($vmid);
4507 return 1 if defined($devices_list->{$deviceid});
4511 die "error on hotplug device
'$deviceid'\n";
4515 sub qemu_devicedelverify {
4516 my ($vmid, $deviceid) = @_;
4518 # need to verify that the device is correctly removed as device_del
4519 # is async and empty return is not reliable
4521 for (my $i = 0; $i <= 5; $i++) {
4522 my $devices_list = vm_devices_list($vmid);
4523 return 1 if !defined($devices_list->{$deviceid});
4527 die "error on hot-unplugging device
'$deviceid'\n";
4530 sub qemu_findorcreatescsihw {
4531 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4533 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4535 my $scsihwid="$controller_prefix$controller";
4536 my $devices_list = vm_devices_list($vmid);
4538 if (!defined($devices_list->{$scsihwid})) {
4539 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4545 sub qemu_deletescsihw {
4546 my ($conf, $vmid, $opt) = @_;
4548 my $device = parse_drive($opt, $conf->{$opt});
4550 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4551 vm_deviceunplug($vmid, $conf, "virtioscsi
$device->{index}");
4555 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4557 my $devices_list = vm_devices_list($vmid);
4558 foreach my $opt (keys %{$devices_list}) {
4559 if (is_valid_drivename($opt)) {
4560 my $drive = parse_drive($opt, $conf->{$opt});
4561 if ($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4567 my $scsihwid="scsihw
$controller";
4569 vm_deviceunplug($vmid, $conf, $scsihwid);
4574 sub qemu_add_pci_bridge {
4575 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4581 print_pci_addr($device, $bridges, $arch, $machine_type);
4583 while (my ($k, $v) = each %$bridges) {
4586 return 1 if !defined($bridgeid) || $bridgeid < 1;
4588 my $bridge = "pci
.$bridgeid";
4589 my $devices_list = vm_devices_list($vmid);
4591 if (!defined($devices_list->{$bridge})) {
4592 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4598 sub qemu_set_link_status {
4599 my ($vmid, $device, $up) = @_;
4601 mon_cmd($vmid, "set_link
", name => $device,
4602 up => $up ? JSON::true : JSON::false);
4605 sub qemu_netdevadd {
4606 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4608 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4609 my %options = split(/[=,]/, $netdev);
4611 if (defined(my $vhost = $options{vhost})) {
4612 $options{vhost} = JSON::boolean(PVE::JSONSchema::parse_boolean($vhost));
4615 if (defined(my $queues = $options{queues})) {
4616 $options{queues} = $queues + 0;
4619 mon_cmd($vmid, "netdev_add
", %options);
4623 sub qemu_netdevdel {
4624 my ($vmid, $deviceid) = @_;
4626 mon_cmd($vmid, "netdev_del
", id => $deviceid);
4629 sub qemu_usb_hotplug {
4630 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4634 # remove the old one first
4635 vm_deviceunplug($vmid, $conf, $deviceid);
4637 # check if xhci controller is necessary and available
4638 my $devicelist = vm_devices_list($vmid);
4640 if (!$devicelist->{xhci}) {
4641 my $pciaddr = print_pci_addr("xhci
", undef, $arch, $machine_type);
4642 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_qemu_xhci_controller($pciaddr));
4646 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type);
4649 sub qemu_cpu_hotplug {
4650 my ($vmid, $conf, $vcpus) = @_;
4652 my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4655 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4656 $sockets = $conf->{sockets} if $conf->{sockets};
4657 my $cores = $conf->{cores} || 1;
4658 my $maxcpus = $sockets * $cores;
4660 $vcpus = $maxcpus if !$vcpus;
4662 die "you can
't add more vcpus than maxcpus\n"
4663 if $vcpus > $maxcpus;
4665 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4667 if ($vcpus < $currentvcpus) {
4669 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4671 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4672 qemu_devicedel($vmid, "cpu$i");
4674 my $currentrunningvcpus = undef;
4676 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4677 last if scalar(@{$currentrunningvcpus}) == $i-1;
4678 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4682 #update conf after each succesfull cpu unplug
4683 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4684 PVE::QemuConfig->write_config($vmid, $conf);
4687 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4693 my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4694 die "vcpus in running vm does not match its configuration\n"
4695 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4697 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4699 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4700 my $cpustr = print_cpu_device($conf, $i);
4701 qemu_deviceadd($vmid, $cpustr);
4704 my $currentrunningvcpus = undef;
4706 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4707 last if scalar(@{$currentrunningvcpus}) == $i;
4708 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4712 #update conf after each succesfull cpu hotplug
4713 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4714 PVE::QemuConfig->write_config($vmid, $conf);
4718 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4719 mon_cmd($vmid, "cpu-add", id => int($i));
4724 sub qemu_block_set_io_throttle {
4725 my ($vmid, $deviceid,
4726 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4727 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4728 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4729 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4731 return if !check_running($vmid) ;
4733 mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4735 bps_rd => int($bps_rd),
4736 bps_wr => int($bps_wr),
4738 iops_rd => int($iops_rd),
4739 iops_wr => int($iops_wr),
4740 bps_max => int($bps_max),
4741 bps_rd_max => int($bps_rd_max),
4742 bps_wr_max => int($bps_wr_max),
4743 iops_max => int($iops_max),
4744 iops_rd_max => int($iops_rd_max),
4745 iops_wr_max => int($iops_wr_max),
4746 bps_max_length => int($bps_max_length),
4747 bps_rd_max_length => int($bps_rd_max_length),
4748 bps_wr_max_length => int($bps_wr_max_length),
4749 iops_max_length => int($iops_max_length),
4750 iops_rd_max_length => int($iops_rd_max_length),
4751 iops_wr_max_length => int($iops_wr_max_length),
4756 sub qemu_block_resize {
4757 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4759 my $running = check_running($vmid);
4761 PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4763 return if !$running;
4765 my $padding = (1024 - $size % 1024) % 1024;
4766 $size = $size + $padding;
4771 device => $deviceid,
4777 sub qemu_volume_snapshot {
4778 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4780 my $running = check_running($vmid);
4782 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4783 mon_cmd($vmid, 'blockdev-snapshot-internal-sync
', device => $deviceid, name => $snap);
4785 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4789 sub qemu_volume_snapshot_delete {
4790 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4792 my $running = check_running($vmid);
4797 my $conf = PVE::QemuConfig->load_config($vmid);
4798 PVE::QemuConfig->foreach_volume($conf, sub {
4799 my ($ds, $drive) = @_;
4800 $running = 1 if $drive->{file} eq $volid;
4804 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4805 mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync
', device => $deviceid, name => $snap);
4807 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4811 sub set_migration_caps {
4812 my ($vmid, $savevm) = @_;
4814 my $qemu_support = eval { mon_cmd($vmid, "query-proxmox-support") };
4816 my $bitmap_prop = $savevm ? 'pbs-dirty-bitmap-savevm
' : 'pbs-dirty-bitmap-migration
';
4817 my $dirty_bitmaps = $qemu_support->{$bitmap_prop} ? 1 : 0;
4822 "auto-converge" => 1,
4824 "x-rdma-pin-all" => 0,
4827 "dirty-bitmaps" => $dirty_bitmaps,
4830 my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4832 for my $supported_capability (@$supported_capabilities) {
4834 capability => $supported_capability->{capability},
4835 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4839 mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4843 my ($conf, $func, @param) = @_;
4847 my $test_volid = sub {
4848 my ($key, $drive, $snapname, $pending) = @_;
4850 my $volid = $drive->{file};
4853 $volhash->{$volid}->{cdrom} //= 1;
4854 $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4856 my $replicate = $drive->{replicate} // 1;
4857 $volhash->{$volid}->{replicate} //= 0;
4858 $volhash->{$volid}->{replicate} = 1 if $replicate;
4860 $volhash->{$volid}->{shared} //= 0;
4861 $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4863 $volhash->{$volid}->{referenced_in_config} //= 0;
4864 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname) && !defined($pending);
4866 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4867 if defined($snapname);
4869 $volhash->{$volid}->{referenced_in_pending} = 1 if defined($pending);
4871 my $size = $drive->{size};
4872 $volhash->{$volid}->{size} //= $size if $size;
4874 $volhash->{$volid}->{is_vmstate} //= 0;
4875 $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate
';
4877 $volhash->{$volid}->{is_tpmstate} //= 0;
4878 $volhash->{$volid}->{is_tpmstate} = 1 if $key eq 'tpmstate0
';
4880 $volhash->{$volid}->{is_unused} //= 0;
4881 $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4883 $volhash->{$volid}->{drivename} = $key if is_valid_drivename($key);
4886 my $include_opts = {
4887 extra_keys => ['vmstate
'],
4888 include_unused => 1,
4891 PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4893 PVE::QemuConfig->foreach_volume_full($conf->{pending}, $include_opts, $test_volid, undef, 1)
4894 if defined($conf->{pending}) && $conf->{pending}->%*;
4896 foreach my $snapname (keys %{$conf->{snapshots}}) {
4897 my $snap = $conf->{snapshots}->{$snapname};
4898 PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4901 foreach my $volid (keys %$volhash) {
4902 &$func($volid, $volhash->{$volid}, @param);
4906 my $fast_plug_option = {
4910 'migrate_downtime
' => 1,
4911 'migrate_speed
' => 1,
4918 'vmstatestorage
' => 1,
4921 for my $opt (keys %$confdesc_cloudinit) {
4922 $fast_plug_option->{$opt} = 1;
4925 # hotplug changes in [PENDING]
4926 # $selection hash can be used to only apply specified options, for
4927 # example: { cores => 1 } (only apply changed 'cores
')
4928 # $errors ref is used to return error messages
4929 sub vmconfig_hotplug_pending {
4930 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4932 my $defaults = load_defaults();
4933 my $arch = get_vm_arch($conf);
4934 my $machine_type = get_vm_machine($conf, undef, $arch);
4936 # commit values which do not have any impact on running VM first
4937 # Note: those option cannot raise errors, we we do not care about
4938 # $selection and always apply them.
4940 my $add_error = sub {
4941 my ($opt, $msg) = @_;
4942 $errors->{$opt} = "hotplug problem - $msg";
4945 my $cloudinit_pending_properties = PVE::QemuServer::cloudinit_pending_properties();
4947 my $cloudinit_record_changed = sub {
4948 my ($conf, $opt, $old, $new) = @_;
4949 return if !$cloudinit_pending_properties->{$opt};
4951 my $ci = ($conf->{cloudinit} //= {});
4953 my $recorded = $ci->{$opt};
4954 my %added = map { $_ => 1 } PVE::Tools::split_list(delete($ci->{added}) // '');
4956 if (defined($new)) {
4957 if (defined($old)) {
4958 # an existing value is being modified
4959 if (defined($recorded)) {
4960 # the value was already not in sync
4961 if ($new eq $recorded) {
4962 # a value is being reverted to the cloud-init state:
4964 delete $added{$opt};
4966 # the value was changed multiple times, do nothing
4968 } elsif ($added{$opt}) {
4969 # the value had been marked as added and is being changed, do nothing
4971 # the value is new, record it:
4975 # a new value is being added
4976 if (defined($recorded)) {
4977 # it was already not in sync
4978 if ($new eq $recorded) {
4979 # a value is being reverted to the cloud-init state:
4981 delete $added{$opt};
4983 # the value had temporarily been removed, do nothing
4985 } elsif ($added{$opt}) {
4986 # the value had been marked as added already, do nothing
4988 # the value is new, add it
4992 } elsif (!defined($old)) {
4993 # a non-existent value is being removed? ignore...
4995 # a value is being deleted
4996 if (defined($recorded)) {
4997 # a value was already recorded, just keep it
4998 } elsif ($added{$opt}) {
4999 # the value was marked as added, remove it
5000 delete $added{$opt};
5002 # a previously unrecorded value is being removed, record the old value:
5007 my $added = join(',', sort keys %added);
5008 $ci->{added} = $added if length($added);
5012 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5013 if ($fast_plug_option->{$opt}) {
5014 my $new = delete $conf->{pending}->{$opt};
5015 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $new);
5016 $conf->{$opt} = $new;
5022 PVE::QemuConfig->write_config($vmid, $conf);
5025 my $ostype = $conf->{ostype};
5026 my $version = extract_version($machine_type, get_running_qemu_version($vmid));
5027 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
5028 my $usb_hotplug = $hotplug_features->{usb}
5029 && min_version($version, 7, 1)
5030 && defined($ostype) && ($ostype eq 'l26
' || windows_version($ostype) > 7);
5032 my $cgroup = PVE::QemuServer::CGroup->new($vmid);
5033 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
5035 foreach my $opt (sort keys %$pending_delete_hash) {
5036 next if $selection && !$selection->{$opt};
5037 my $force = $pending_delete_hash->{$opt}->{force};
5039 if ($opt eq 'hotplug
') {
5040 die "skip\n" if ($conf->{hotplug} =~ /memory/);
5041 } elsif ($opt eq 'tablet
') {
5042 die "skip\n" if !$hotplug_features->{usb};
5043 if ($defaults->{tablet}) {
5044 vm_deviceplug($storecfg, $conf, $vmid, 'tablet
', $arch, $machine_type);
5045 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard
', $arch, $machine_type)
5046 if $arch eq 'aarch64
';
5048 vm_deviceunplug($vmid, $conf, 'tablet
');
5049 vm_deviceunplug($vmid, $conf, 'keyboard
') if $arch eq 'aarch64
';
5051 } elsif ($opt =~ m/^usb(\d+)$/) {
5053 die "skip\n" if !$usb_hotplug;
5054 vm_deviceunplug($vmid, $conf, "usbredirdev$index"); # if it's a spice port
5055 vm_deviceunplug
($vmid, $conf, $opt);
5056 } elsif ($opt eq 'vcpus') {
5057 die "skip\n" if !$hotplug_features->{cpu
};
5058 qemu_cpu_hotplug
($vmid, $conf, undef);
5059 } elsif ($opt eq 'balloon') {
5060 # enable balloon device is not hotpluggable
5061 die "skip\n" if defined($conf->{balloon
}) && $conf->{balloon
} == 0;
5062 # here we reset the ballooning value to memory
5063 my $balloon = $conf->{memory
} || $defaults->{memory
};
5064 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5065 } elsif ($fast_plug_option->{$opt}) {
5067 } elsif ($opt =~ m/^net(\d+)$/) {
5068 die "skip\n" if !$hotplug_features->{network
};
5069 vm_deviceunplug
($vmid, $conf, $opt);
5070 } elsif (is_valid_drivename
($opt)) {
5071 die "skip\n" if !$hotplug_features->{disk
} || $opt =~ m/(ide|sata)(\d+)/;
5072 vm_deviceunplug
($vmid, $conf, $opt);
5073 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5074 } elsif ($opt =~ m/^memory$/) {
5075 die "skip\n" if !$hotplug_features->{memory
};
5076 PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults);
5077 } elsif ($opt eq 'cpuunits') {
5078 $cgroup->change_cpu_shares(undef);
5079 } elsif ($opt eq 'cpulimit') {
5080 $cgroup->change_cpu_quota(undef, undef); # reset, cgroup module can better decide values
5086 &$add_error($opt, $err) if $err ne "skip\n";
5088 my $old = delete $conf->{$opt};
5089 $cloudinit_record_changed->($conf, $opt, $old, undef);
5090 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5095 foreach my $opt (keys %{$conf->{pending
}}) {
5096 next if $selection && !$selection->{$opt};
5097 my $value = $conf->{pending
}->{$opt};
5099 if ($opt eq 'hotplug') {
5100 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug
} =~ /memory/);
5101 } elsif ($opt eq 'tablet') {
5102 die "skip\n" if !$hotplug_features->{usb
};
5104 vm_deviceplug
($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5105 vm_deviceplug
($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5106 if $arch eq 'aarch64';
5107 } elsif ($value == 0) {
5108 vm_deviceunplug
($vmid, $conf, 'tablet');
5109 vm_deviceunplug
($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5111 } elsif ($opt =~ m/^usb(\d+)$/) {
5113 die "skip\n" if !$usb_hotplug;
5114 my $d = eval { parse_property_string
('pve-qm-usb', $value) };
5116 if ($d->{host
} =~ m/^spice$/i) {
5117 $id = "usbredirdev$index";
5119 qemu_usb_hotplug
($storecfg, $conf, $vmid, $id, $d, $arch, $machine_type);
5120 } elsif ($opt eq 'vcpus') {
5121 die "skip\n" if !$hotplug_features->{cpu
};
5122 qemu_cpu_hotplug
($vmid, $conf, $value);
5123 } elsif ($opt eq 'balloon') {
5124 # enable/disable balloning device is not hotpluggable
5125 my $old_balloon_enabled = !!(!defined($conf->{balloon
}) || $conf->{balloon
});
5126 my $new_balloon_enabled = !!(!defined($conf->{pending
}->{balloon
}) || $conf->{pending
}->{balloon
});
5127 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5129 # allow manual ballooning if shares is set to zero
5130 if ((defined($conf->{shares
}) && ($conf->{shares
} == 0))) {
5131 my $balloon = $conf->{pending
}->{balloon
} || $conf->{memory
} || $defaults->{memory
};
5132 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5134 } elsif ($opt =~ m/^net(\d+)$/) {
5135 # some changes can be done without hotplug
5136 vmconfig_update_net
($storecfg, $conf, $hotplug_features->{network
},
5137 $vmid, $opt, $value, $arch, $machine_type);
5138 } elsif (is_valid_drivename
($opt)) {
5139 die "skip\n" if $opt eq 'efidisk0' || $opt eq 'tpmstate0';
5140 # some changes can be done without hotplug
5141 my $drive = parse_drive
($opt, $value);
5142 if (drive_is_cloudinit
($drive)) {
5143 $cloudinit_opt = [$opt, $drive];
5144 # apply all the other changes first, then generate the cloudinit disk
5147 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5148 $vmid, $opt, $value, $arch, $machine_type);
5149 } elsif ($opt =~ m/^memory$/) { #dimms
5150 die "skip\n" if !$hotplug_features->{memory
};
5151 $value = PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults, $value);
5152 } elsif ($opt eq 'cpuunits') {
5153 my $new_cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{pending
}->{$opt}); #clamp
5154 $cgroup->change_cpu_shares($new_cpuunits);
5155 } elsif ($opt eq 'cpulimit') {
5156 my $cpulimit = $conf->{pending
}->{$opt} == 0 ?
-1 : int($conf->{pending
}->{$opt} * 100000);
5157 $cgroup->change_cpu_quota($cpulimit, 100000);
5158 } elsif ($opt eq 'agent') {
5159 vmconfig_update_agent
($conf, $opt, $value);
5161 die "skip\n"; # skip non-hot-pluggable options
5165 &$add_error($opt, $err) if $err ne "skip\n";
5167 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $value);
5168 $conf->{$opt} = $value;
5169 delete $conf->{pending
}->{$opt};
5173 if (defined($cloudinit_opt)) {
5174 my ($opt, $drive) = @$cloudinit_opt;
5175 my $value = $conf->{pending
}->{$opt};
5177 my $temp = {%$conf, $opt => $value};
5178 PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($temp, $vmid);
5179 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5180 $vmid, $opt, $value, $arch, $machine_type);
5183 &$add_error($opt, $err) if $err ne "skip\n";
5185 $conf->{$opt} = $value;
5186 delete $conf->{pending
}->{$opt};
5190 # unplug xhci controller if no usb device is left
5193 for (my $i = 0; $i < $PVE::QemuServer
::USB
::MAX_USB_DEVICES
; $i++) {
5194 next if !defined($conf->{"usb$i"});
5199 vm_deviceunplug
($vmid, $conf, 'xhci');
5203 PVE
::QemuConfig-
>write_config($vmid, $conf);
5205 if ($hotplug_features->{cloudinit
} && PVE
::QemuServer
::Cloudinit
::has_changes
($conf)) {
5206 PVE
::QemuServer
::vmconfig_update_cloudinit_drive
($storecfg, $conf, $vmid);
5210 sub try_deallocate_drive
{
5211 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5213 if (($force || $key =~ /^unused/) && !drive_is_cdrom
($drive, 1)) {
5214 my $volid = $drive->{file
};
5215 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
5216 my $sid = PVE
::Storage
::parse_volume_id
($volid);
5217 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5219 # check if the disk is really unused
5220 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5221 if PVE
::QemuServer
::Drive
::is_volume_in_use
($storecfg, $conf, $key, $volid);
5222 PVE
::Storage
::vdisk_free
($storecfg, $volid);
5225 # If vm is not owner of this disk remove from config
5233 sub vmconfig_delete_or_detach_drive
{
5234 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5236 my $drive = parse_drive
($opt, $conf->{$opt});
5238 my $rpcenv = PVE
::RPCEnvironment
::get
();
5239 my $authuser = $rpcenv->get_user();
5242 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5243 try_deallocate_drive
($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5245 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $drive);
5251 sub vmconfig_apply_pending
{
5252 my ($vmid, $conf, $storecfg, $errors, $skip_cloud_init) = @_;
5254 return if !scalar(keys %{$conf->{pending
}});
5256 my $add_apply_error = sub {
5257 my ($opt, $msg) = @_;
5258 my $err_msg = "unable to apply pending change $opt : $msg";
5259 $errors->{$opt} = $err_msg;
5265 my $pending_delete_hash = PVE
::QemuConfig-
>parse_pending_delete($conf->{pending
}->{delete});
5266 foreach my $opt (sort keys %$pending_delete_hash) {
5267 my $force = $pending_delete_hash->{$opt}->{force
};
5269 if ($opt =~ m/^unused/) {
5270 die "internal error";
5271 } elsif (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5272 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5276 $add_apply_error->($opt, $err);
5278 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5279 delete $conf->{$opt};
5283 PVE
::QemuConfig-
>cleanup_pending($conf);
5285 my $generate_cloudinit = $skip_cloud_init ?
0 : undef;
5287 foreach my $opt (keys %{$conf->{pending
}}) { # add/change
5288 next if $opt eq 'delete'; # just to be sure
5290 if (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5291 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, parse_drive
($opt, $conf->{$opt}))
5295 $add_apply_error->($opt, $err);
5298 if (is_valid_drivename
($opt)) {
5299 my $drive = parse_drive
($opt, $conf->{pending
}->{$opt});
5300 $generate_cloudinit //= 1 if drive_is_cloudinit
($drive);
5303 $conf->{$opt} = delete $conf->{pending
}->{$opt};
5307 # write all changes at once to avoid unnecessary i/o
5308 PVE
::QemuConfig-
>write_config($vmid, $conf);
5309 if ($generate_cloudinit) {
5310 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5311 # After successful generation and if there were changes to be applied, update the
5312 # config to drop the {cloudinit} entry.
5313 PVE
::QemuConfig-
>write_config($vmid, $conf);
5318 sub vmconfig_update_net
{
5319 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5321 my $newnet = parse_net
($value);
5323 if ($conf->{$opt}) {
5324 my $oldnet = parse_net
($conf->{$opt});
5326 if (safe_string_ne
($oldnet->{model
}, $newnet->{model
}) ||
5327 safe_string_ne
($oldnet->{macaddr
}, $newnet->{macaddr
}) ||
5328 safe_num_ne
($oldnet->{queues
}, $newnet->{queues
}) ||
5329 safe_num_ne
($oldnet->{mtu
}, $newnet->{mtu
}) ||
5330 !($newnet->{bridge
} && $oldnet->{bridge
})) { # bridge/nat mode change
5332 # for non online change, we try to hot-unplug
5333 die "skip\n" if !$hotplug;
5334 vm_deviceunplug
($vmid, $conf, $opt);
5337 die "internal error" if $opt !~ m/net(\d+)/;
5338 my $iface = "tap${vmid}i$1";
5340 if (safe_string_ne
($oldnet->{bridge
}, $newnet->{bridge
}) ||
5341 safe_num_ne
($oldnet->{tag
}, $newnet->{tag
}) ||
5342 safe_string_ne
($oldnet->{trunks
}, $newnet->{trunks
}) ||
5343 safe_num_ne
($oldnet->{firewall
}, $newnet->{firewall
})) {
5344 PVE
::Network
::tap_unplug
($iface);
5347 PVE
::Network
::SDN
::Zones
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5349 PVE
::Network
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5351 } elsif (safe_num_ne
($oldnet->{rate
}, $newnet->{rate
})) {
5352 # Rate can be applied on its own but any change above needs to
5353 # include the rate in tap_plug since OVS resets everything.
5354 PVE
::Network
::tap_rate_limit
($iface, $newnet->{rate
});
5357 if (safe_string_ne
($oldnet->{link_down
}, $newnet->{link_down
})) {
5358 qemu_set_link_status
($vmid, $opt, !$newnet->{link_down
});
5366 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5372 sub vmconfig_update_agent
{
5373 my ($conf, $opt, $value) = @_;
5375 die "skip\n" if !$conf->{$opt};
5377 my $hotplug_options = { fstrim_cloned_disks
=> 1 };
5379 my $old_agent = parse_guest_agent
($conf);
5380 my $agent = parse_guest_agent
({$opt => $value});
5382 for my $option (keys %$agent) { # added/changed options
5383 next if defined($hotplug_options->{$option});
5384 die "skip\n" if safe_string_ne
($agent->{$option}, $old_agent->{$option});
5387 for my $option (keys %$old_agent) { # removed options
5388 next if defined($hotplug_options->{$option});
5389 die "skip\n" if safe_string_ne
($old_agent->{$option}, $agent->{$option});
5392 return; # either no actual change (e.g., format string reordered) or just hotpluggable changes
5395 sub vmconfig_update_disk
{
5396 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5398 my $drive = parse_drive
($opt, $value);
5400 if ($conf->{$opt} && (my $old_drive = parse_drive
($opt, $conf->{$opt}))) {
5401 my $media = $drive->{media
} || 'disk';
5402 my $oldmedia = $old_drive->{media
} || 'disk';
5403 die "unable to change media type\n" if $media ne $oldmedia;
5405 if (!drive_is_cdrom
($old_drive)) {
5407 if ($drive->{file
} ne $old_drive->{file
}) {
5409 die "skip\n" if !$hotplug;
5411 # unplug and register as unused
5412 vm_deviceunplug
($vmid, $conf, $opt);
5413 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive)
5416 # update existing disk
5418 # skip non hotpluggable value
5419 if (safe_string_ne
($drive->{aio
}, $old_drive->{aio
}) ||
5420 safe_string_ne
($drive->{discard
}, $old_drive->{discard
}) ||
5421 safe_string_ne
($drive->{iothread
}, $old_drive->{iothread
}) ||
5422 safe_string_ne
($drive->{queues
}, $old_drive->{queues
}) ||
5423 safe_string_ne
($drive->{cache
}, $old_drive->{cache
}) ||
5424 safe_string_ne
($drive->{ssd
}, $old_drive->{ssd
}) ||
5425 safe_string_ne
($drive->{ro
}, $old_drive->{ro
})) {
5430 if (safe_num_ne
($drive->{mbps
}, $old_drive->{mbps
}) ||
5431 safe_num_ne
($drive->{mbps_rd
}, $old_drive->{mbps_rd
}) ||
5432 safe_num_ne
($drive->{mbps_wr
}, $old_drive->{mbps_wr
}) ||
5433 safe_num_ne
($drive->{iops
}, $old_drive->{iops
}) ||
5434 safe_num_ne
($drive->{iops_rd
}, $old_drive->{iops_rd
}) ||
5435 safe_num_ne
($drive->{iops_wr
}, $old_drive->{iops_wr
}) ||
5436 safe_num_ne
($drive->{mbps_max
}, $old_drive->{mbps_max
}) ||
5437 safe_num_ne
($drive->{mbps_rd_max
}, $old_drive->{mbps_rd_max
}) ||
5438 safe_num_ne
($drive->{mbps_wr_max
}, $old_drive->{mbps_wr_max
}) ||
5439 safe_num_ne
($drive->{iops_max
}, $old_drive->{iops_max
}) ||
5440 safe_num_ne
($drive->{iops_rd_max
}, $old_drive->{iops_rd_max
}) ||
5441 safe_num_ne
($drive->{iops_wr_max
}, $old_drive->{iops_wr_max
}) ||
5442 safe_num_ne
($drive->{bps_max_length
}, $old_drive->{bps_max_length
}) ||
5443 safe_num_ne
($drive->{bps_rd_max_length
}, $old_drive->{bps_rd_max_length
}) ||
5444 safe_num_ne
($drive->{bps_wr_max_length
}, $old_drive->{bps_wr_max_length
}) ||
5445 safe_num_ne
($drive->{iops_max_length
}, $old_drive->{iops_max_length
}) ||
5446 safe_num_ne
($drive->{iops_rd_max_length
}, $old_drive->{iops_rd_max_length
}) ||
5447 safe_num_ne
($drive->{iops_wr_max_length
}, $old_drive->{iops_wr_max_length
})) {
5449 qemu_block_set_io_throttle
(
5451 ($drive->{mbps
} || 0)*1024*1024,
5452 ($drive->{mbps_rd
} || 0)*1024*1024,
5453 ($drive->{mbps_wr
} || 0)*1024*1024,
5454 $drive->{iops
} || 0,
5455 $drive->{iops_rd
} || 0,
5456 $drive->{iops_wr
} || 0,
5457 ($drive->{mbps_max
} || 0)*1024*1024,
5458 ($drive->{mbps_rd_max
} || 0)*1024*1024,
5459 ($drive->{mbps_wr_max
} || 0)*1024*1024,
5460 $drive->{iops_max
} || 0,
5461 $drive->{iops_rd_max
} || 0,
5462 $drive->{iops_wr_max
} || 0,
5463 $drive->{bps_max_length
} || 1,
5464 $drive->{bps_rd_max_length
} || 1,
5465 $drive->{bps_wr_max_length
} || 1,
5466 $drive->{iops_max_length
} || 1,
5467 $drive->{iops_rd_max_length
} || 1,
5468 $drive->{iops_wr_max_length
} || 1,
5478 if ($drive->{file
} eq 'none') {
5479 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5480 if (drive_is_cloudinit
($old_drive)) {
5481 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive);
5484 my $path = get_iso_path
($storecfg, $vmid, $drive->{file
});
5486 # force eject if locked
5487 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5490 mon_cmd
($vmid, "blockdev-change-medium",
5491 id
=> "$opt", filename
=> "$path");
5499 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5501 PVE
::Storage
::activate_volumes
($storecfg, [$drive->{file
}]) if $drive->{file
} !~ m
|^/dev/.+|;
5502 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5505 sub vmconfig_update_cloudinit_drive
{
5506 my ($storecfg, $conf, $vmid) = @_;
5508 my $cloudinit_ds = undef;
5509 my $cloudinit_drive = undef;
5511 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5512 my ($ds, $drive) = @_;
5513 if (PVE
::QemuServer
::drive_is_cloudinit
($drive)) {
5514 $cloudinit_ds = $ds;
5515 $cloudinit_drive = $drive;
5519 return if !$cloudinit_drive;
5521 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5522 PVE
::QemuConfig-
>write_config($vmid, $conf);
5525 my $running = PVE
::QemuServer
::check_running
($vmid);
5528 my $path = PVE
::Storage
::path
($storecfg, $cloudinit_drive->{file
});
5530 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$cloudinit_ds");
5531 mon_cmd
($vmid, "blockdev-change-medium", id
=> "$cloudinit_ds", filename
=> "$path");
5536 # called in locked context by incoming migration
5537 sub vm_migrate_get_nbd_disks
{
5538 my ($storecfg, $conf, $replicated_volumes) = @_;
5540 my $local_volumes = {};
5541 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5542 my ($ds, $drive) = @_;
5544 return if drive_is_cdrom
($drive);
5545 return if $ds eq 'tpmstate0';
5547 my $volid = $drive->{file
};
5551 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
5553 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5554 return if $scfg->{shared
};
5556 # replicated disks re-use existing state via bitmap
5557 my $use_existing = $replicated_volumes->{$volid} ?
1 : 0;
5558 $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
5560 return $local_volumes;
5563 # called in locked context by incoming migration
5564 sub vm_migrate_alloc_nbd_disks
{
5565 my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
5568 foreach my $opt (sort keys %$source_volumes) {
5569 my ($volid, $storeid, $volname, $drive, $use_existing, $format) = @{$source_volumes->{$opt}};
5571 if ($use_existing) {
5572 $nbd->{$opt}->{drivestr
} = print_drive
($drive);
5573 $nbd->{$opt}->{volid
} = $volid;
5574 $nbd->{$opt}->{replicated
} = 1;
5578 # storage mapping + volname = regular migration
5579 # storage mapping + format = remote migration
5580 # order of precedence, filtered by whether storage supports it:
5581 # 1. explicit requested format
5582 # 2. format of current volume
5583 # 3. default format of storage
5584 if (!$storagemap->{identity
}) {
5585 $storeid = PVE
::JSONSchema
::map_id
($storagemap, $storeid);
5586 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
5587 if (!$format || !grep { $format eq $_ } @$validFormats) {
5589 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5590 my $fileFormat = qemu_img_format
($scfg, $volname);
5591 $format = $fileFormat
5592 if grep { $fileFormat eq $_ } @$validFormats;
5594 $format //= $defFormat;
5597 # can't happen for remote migration, so $volname is always defined
5598 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5599 $format = qemu_img_format
($scfg, $volname);
5602 my $size = $drive->{size
} / 1024;
5603 my $newvolid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $format, undef, $size);
5604 my $newdrive = $drive;
5605 $newdrive->{format
} = $format;
5606 $newdrive->{file
} = $newvolid;
5607 my $drivestr = print_drive
($newdrive);
5608 $nbd->{$opt}->{drivestr
} = $drivestr;
5609 $nbd->{$opt}->{volid
} = $newvolid;
5615 # see vm_start_nolock for parameters, additionally:
5617 # storagemap = parsed storage map for allocating NBD disks
5619 my ($storecfg, $vmid, $params, $migrate_opts) = @_;
5621 return PVE
::QemuConfig-
>lock_config($vmid, sub {
5622 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migrate_opts->{migratedfrom
});
5624 die "you can't start a vm if it's a template\n"
5625 if !$params->{skiptemplate
} && PVE
::QemuConfig-
>is_template($conf);
5627 my $has_suspended_lock = PVE
::QemuConfig-
>has_lock($conf, 'suspended');
5628 my $has_backup_lock = PVE
::QemuConfig-
>has_lock($conf, 'backup');
5630 my $running = check_running
($vmid, undef, $migrate_opts->{migratedfrom
});
5632 if ($has_backup_lock && $running) {
5633 # a backup is currently running, attempt to start the guest in the
5634 # existing QEMU instance
5635 return vm_resume
($vmid);
5638 PVE
::QemuConfig-
>check_lock($conf)
5639 if !($params->{skiplock
} || $has_suspended_lock);
5641 $params->{resume
} = $has_suspended_lock || defined($conf->{vmstate
});
5643 die "VM $vmid already running\n" if $running;
5645 if (my $storagemap = $migrate_opts->{storagemap
}) {
5646 my $replicated = $migrate_opts->{replicated_volumes
};
5647 my $disks = vm_migrate_get_nbd_disks
($storecfg, $conf, $replicated);
5648 $migrate_opts->{nbd
} = vm_migrate_alloc_nbd_disks
($storecfg, $vmid, $disks, $storagemap);
5650 foreach my $opt (keys %{$migrate_opts->{nbd
}}) {
5651 $conf->{$opt} = $migrate_opts->{nbd
}->{$opt}->{drivestr
};
5655 return vm_start_nolock
($storecfg, $vmid, $conf, $params, $migrate_opts);
5661 # statefile => 'tcp', 'unix' for migration or path/volid for RAM state
5662 # skiplock => 0/1, skip checking for config lock
5663 # skiptemplate => 0/1, skip checking whether VM is template
5664 # forcemachine => to force QEMU machine (rollback/migration)
5665 # forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
5666 # timeout => in seconds
5667 # paused => start VM in paused state (backup)
5668 # resume => resume from hibernation
5679 # nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
5680 # migratedfrom => source node
5681 # spice_ticket => used for spice migration, passed via tunnel/stdin
5682 # network => CIDR of migration network
5683 # type => secure/insecure - tunnel over encrypted connection or plain-text
5684 # nbd_proto_version => int, 0 for TCP, 1 for UNIX
5685 # replicated_volumes => which volids should be re-used with bitmaps for nbd migration
5686 # offline_volumes => new volids of offline migrated disks like tpmstate and cloudinit, not yet
5687 # contained in config
5688 sub vm_start_nolock
{
5689 my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
5691 my $statefile = $params->{statefile
};
5692 my $resume = $params->{resume
};
5694 my $migratedfrom = $migrate_opts->{migratedfrom
};
5695 my $migration_type = $migrate_opts->{type
};
5699 # clean up leftover reboot request files
5700 eval { clear_reboot_request
($vmid); };
5703 if (!$statefile && scalar(keys %{$conf->{pending
}})) {
5704 vmconfig_apply_pending
($vmid, $conf, $storecfg);
5705 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5708 # don't regenerate the ISO if the VM is started as part of a live migration
5709 # this way we can reuse the old ISO with the correct config
5710 if (!$migratedfrom) {
5711 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5712 # FIXME: apply_cloudinit_config updates $conf in this case, and it would only drop
5713 # $conf->{cloudinit}, so we could just not do this?
5714 # But we do it above, so for now let's be consistent.
5715 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5719 # override offline migrated volumes, conf is out of date still
5720 if (my $offline_volumes = $migrate_opts->{offline_volumes
}) {
5721 for my $key (sort keys $offline_volumes->%*) {
5722 my $parsed = parse_drive
($key, $conf->{$key});
5723 $parsed->{file
} = $offline_volumes->{$key};
5724 $conf->{$key} = print_drive
($parsed);
5728 my $defaults = load_defaults
();
5730 # set environment variable useful inside network script
5731 # for remote migration the config is available on the target node!
5732 if (!$migrate_opts->{remote_node
}) {
5733 $ENV{PVE_MIGRATED_FROM
} = $migratedfrom;
5736 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-start', 1);
5738 my $forcemachine = $params->{forcemachine
};
5739 my $forcecpu = $params->{forcecpu
};
5741 # enforce machine and CPU type on suspended vm to ensure HW compatibility
5742 $forcemachine = $conf->{runningmachine
};
5743 $forcecpu = $conf->{runningcpu
};
5744 print "Resuming suspended VM\n";
5747 my ($cmd, $vollist, $spice_port, $pci_devices) = config_to_command
($storecfg, $vmid,
5748 $conf, $defaults, $forcemachine, $forcecpu, $params->{'pbs-backing'});
5751 my $get_migration_ip = sub {
5752 my ($nodename) = @_;
5754 return $migration_ip if defined($migration_ip);
5756 my $cidr = $migrate_opts->{network
};
5758 if (!defined($cidr)) {
5759 my $dc_conf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5760 $cidr = $dc_conf->{migration
}->{network
};
5763 if (defined($cidr)) {
5764 my $ips = PVE
::Network
::get_local_ip_from_cidr
($cidr);
5766 die "could not get IP: no address configured on local " .
5767 "node for network '$cidr'\n" if scalar(@$ips) == 0;
5769 die "could not get IP: multiple addresses configured on local " .
5770 "node for network '$cidr'\n" if scalar(@$ips) > 1;
5772 $migration_ip = @$ips[0];
5775 $migration_ip = PVE
::Cluster
::remote_node_ip
($nodename, 1)
5776 if !defined($migration_ip);
5778 return $migration_ip;
5782 if ($statefile eq 'tcp') {
5783 my $migrate = $res->{migrate
} = { proto
=> 'tcp' };
5784 $migrate->{addr
} = "localhost";
5785 my $datacenterconf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5786 my $nodename = nodename
();
5788 if (!defined($migration_type)) {
5789 if (defined($datacenterconf->{migration
}->{type
})) {
5790 $migration_type = $datacenterconf->{migration
}->{type
};
5792 $migration_type = 'secure';
5796 if ($migration_type eq 'insecure') {
5797 $migrate->{addr
} = $get_migration_ip->($nodename);
5798 $migrate->{addr
} = "[$migrate->{addr}]" if Net
::IP
::ip_is_ipv6
($migrate->{addr
});
5801 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5802 $migrate->{port
} = PVE
::Tools
::next_migrate_port
($pfamily);
5803 $migrate->{uri
} = "tcp:$migrate->{addr}:$migrate->{port}";
5804 push @$cmd, '-incoming', $migrate->{uri
};
5807 } elsif ($statefile eq 'unix') {
5808 # should be default for secure migrations as a ssh TCP forward
5809 # tunnel is not deterministic reliable ready and fails regurarly
5810 # to set up in time, so use UNIX socket forwards
5811 my $migrate = $res->{migrate
} = { proto
=> 'unix' };
5812 $migrate->{addr
} = "/run/qemu-server/$vmid.migrate";
5813 unlink $migrate->{addr
};
5815 $migrate->{uri
} = "unix:$migrate->{addr}";
5816 push @$cmd, '-incoming', $migrate->{uri
};
5819 } elsif (-e
$statefile) {
5820 push @$cmd, '-loadstate', $statefile;
5822 my $statepath = PVE
::Storage
::path
($storecfg, $statefile);
5823 push @$vollist, $statefile;
5824 push @$cmd, '-loadstate', $statepath;
5826 } elsif ($params->{paused
}) {
5830 my $start_timeout = $params->{timeout
} // config_aware_timeout
($conf, $resume);
5832 my $pci_reserve_list = [];
5833 for my $device (values $pci_devices->%*) {
5834 next if $device->{mdev
}; # we don't reserve for mdev devices
5835 push $pci_reserve_list->@*, map { $_->{id
} } $device->{ids
}->@*;
5838 # reserve all PCI IDs before actually doing anything with them
5839 PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_reserve_list, $vmid, $start_timeout);
5843 for my $id (sort keys %$pci_devices) {
5844 my $d = $pci_devices->{$id};
5845 my ($index) = ($id =~ m/^hostpci(\d+)$/);
5848 for my $dev ($d->{ids
}->@*) {
5849 my $info = eval { PVE
::QemuServer
::PCI
::prepare_pci_device
($vmid, $dev->{id
}, $index, $d->{mdev
}) };
5852 $chosen_mdev = $info;
5853 last if $chosen_mdev; # if successful, we're done
5859 next if !$d->{mdev
};
5860 die "could not create mediated device\n" if !defined($chosen_mdev);
5862 # nvidia grid needs the uuid of the mdev as qemu parameter
5863 if (!defined($uuid) && $chosen_mdev->{vendor
} =~ m/^(0x)?10de$/) {
5864 if (defined($conf->{smbios1
})) {
5865 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
5866 $uuid = $smbios_conf->{uuid
} if defined($smbios_conf->{uuid
});
5868 $uuid = PVE
::QemuServer
::PCI
::generate_mdev_uuid
($vmid, $index) if !defined($uuid);
5871 push @$cmd, '-uuid', $uuid if defined($uuid);
5874 eval { cleanup_pci_devices
($vmid, $conf) };
5879 PVE
::Storage
::activate_volumes
($storecfg, $vollist);
5882 my %silence_std_outs = (outfunc
=> sub {}, errfunc
=> sub {});
5883 eval { run_command
(['/bin/systemctl', 'reset-failed', "$vmid.scope"], %silence_std_outs) };
5884 eval { run_command
(['/bin/systemctl', 'stop', "$vmid.scope"], %silence_std_outs) };
5885 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5886 # timeout should be more than enough here...
5887 PVE
::Systemd
::wait_for_unit_removed
("$vmid.scope", 20);
5889 my $cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{cpuunits
});
5892 timeout
=> $statefile ?
undef : $start_timeout,
5897 # when migrating, prefix QEMU output so other side can pick up any
5898 # errors that might occur and show the user
5899 if ($migratedfrom) {
5900 $run_params{quiet
} = 1;
5901 $run_params{logfunc
} = sub { print "QEMU: $_[0]\n" };
5904 my %systemd_properties = (
5905 Slice
=> 'qemu.slice',
5906 KillMode
=> 'process',
5908 TimeoutStopUSec
=> ULONG_MAX
, # infinity
5911 if (PVE
::CGroup
::cgroup_mode
() == 2) {
5912 $systemd_properties{CPUWeight
} = $cpuunits;
5914 $systemd_properties{CPUShares
} = $cpuunits;
5917 if (my $cpulimit = $conf->{cpulimit
}) {
5918 $systemd_properties{CPUQuota
} = int($cpulimit * 100);
5920 $systemd_properties{timeout
} = 10 if $statefile; # setting up the scope shoul be quick
5922 my $run_qemu = sub {
5923 PVE
::Tools
::run_fork
sub {
5924 PVE
::Systemd
::enter_systemd_scope
($vmid, "Proxmox VE VM $vmid", %systemd_properties);
5927 if (my $tpm = $conf->{tpmstate0
}) {
5928 # start the TPM emulator so QEMU can connect on start
5929 $tpmpid = start_swtpm
($storecfg, $vmid, $tpm, $migratedfrom);
5932 my $exitcode = run_command
($cmd, %run_params);
5935 warn "stopping swtpm instance (pid $tpmpid) due to QEMU startup error\n";
5936 kill 'TERM', $tpmpid;
5938 die "QEMU exited with code $exitcode\n";
5943 if ($conf->{hugepages
}) {
5946 my $hotplug_features =
5947 parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
5948 my $hugepages_topology =
5949 PVE
::QemuServer
::Memory
::hugepages_topology
($conf, $hotplug_features->{memory
});
5951 my $hugepages_host_topology = PVE
::QemuServer
::Memory
::hugepages_host_topology
();
5953 PVE
::QemuServer
::Memory
::hugepages_mount
();
5954 PVE
::QemuServer
::Memory
::hugepages_allocate
($hugepages_topology, $hugepages_host_topology);
5956 eval { $run_qemu->() };
5958 PVE
::QemuServer
::Memory
::hugepages_reset
($hugepages_host_topology)
5959 if !$conf->{keephugepages
};
5963 PVE
::QemuServer
::Memory
::hugepages_pre_deallocate
($hugepages_topology)
5964 if !$conf->{keephugepages
};
5966 eval { PVE
::QemuServer
::Memory
::hugepages_update_locked
($code); };
5969 eval { $run_qemu->() };
5973 # deactivate volumes if start fails
5974 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
5976 eval { cleanup_pci_devices
($vmid, $conf) };
5979 die "start failed: $err";
5982 # re-reserve all PCI IDs now that we can know the actual VM PID
5983 my $pid = PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
5984 eval { PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_reserve_list, $vmid, undef, $pid) };
5987 if (defined($res->{migrate
})) {
5988 print "migration listens on $res->{migrate}->{uri}\n";
5989 } elsif ($statefile) {
5990 eval { mon_cmd
($vmid, "cont"); };
5994 #start nbd server for storage migration
5995 if (my $nbd = $migrate_opts->{nbd
}) {
5996 my $nbd_protocol_version = $migrate_opts->{nbd_proto_version
} // 0;
5998 my $migrate_storage_uri;
5999 # nbd_protocol_version > 0 for unix socket support
6000 if ($nbd_protocol_version > 0 && ($migration_type eq 'secure' || $migration_type eq 'websocket')) {
6001 my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
6002 mon_cmd
($vmid, "nbd-server-start", addr
=> { type
=> 'unix', data
=> { path
=> $socket_path } } );
6003 $migrate_storage_uri = "nbd:unix:$socket_path";
6004 $res->{migrate
}->{unix_sockets
} = [$socket_path];
6006 my $nodename = nodename
();
6007 my $localip = $get_migration_ip->($nodename);
6008 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
6009 my $storage_migrate_port = PVE
::Tools
::next_migrate_port
($pfamily);
6011 mon_cmd
($vmid, "nbd-server-start", addr
=> {
6014 host
=> "${localip}",
6015 port
=> "${storage_migrate_port}",
6018 $localip = "[$localip]" if Net
::IP
::ip_is_ipv6
($localip);
6019 $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
6022 my $block_info = mon_cmd
($vmid, "query-block");
6023 $block_info = { map { $_->{device
} => $_ } $block_info->@* };
6025 foreach my $opt (sort keys %$nbd) {
6026 my $drivestr = $nbd->{$opt}->{drivestr
};
6027 my $volid = $nbd->{$opt}->{volid
};
6029 my $block_node = $block_info->{"drive-$opt"}->{inserted
}->{'node-name'};
6035 'node-name' => $block_node,
6036 writable
=> JSON
::true
,
6038 name
=> "drive-$opt", # NBD export name
6041 my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
6042 print "storage migration listens on $nbd_uri volume:$drivestr\n";
6043 print "re-using replicated volume: $opt - $volid\n"
6044 if $nbd->{$opt}->{replicated
};
6046 $res->{drives
}->{$opt} = $nbd->{$opt};
6047 $res->{drives
}->{$opt}->{nbd_uri
} = $nbd_uri;
6051 if ($migratedfrom) {
6053 set_migration_caps
($vmid);
6058 print "spice listens on port $spice_port\n";
6059 $res->{spice_port
} = $spice_port;
6060 if ($migrate_opts->{spice_ticket
}) {
6061 mon_cmd
($vmid, "set_password", protocol
=> 'spice', password
=>
6062 $migrate_opts->{spice_ticket
});
6063 mon_cmd
($vmid, "expire_password", protocol
=> 'spice', time => "+30");
6068 mon_cmd
($vmid, "balloon", value
=> $conf->{balloon
}*1024*1024)
6069 if !$statefile && $conf->{balloon
};
6071 foreach my $opt (keys %$conf) {
6072 next if $opt !~ m/^net\d+$/;
6073 my $nicconf = parse_net
($conf->{$opt});
6074 qemu_set_link_status
($vmid, $opt, 0) if $nicconf->{link_down
};
6076 add_nets_bridge_fdb
($conf, $vmid);
6079 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
6084 path
=> "machine/peripheral/balloon0",
6085 property
=> "guest-stats-polling-interval",
6089 log_warn
("could not set polling interval for ballooning - $@") if $@;
6093 print "Resumed VM, removing state\n";
6094 if (my $vmstate = $conf->{vmstate
}) {
6095 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6096 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6098 delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
6099 PVE
::QemuConfig-
>write_config($vmid, $conf);
6102 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'post-start');
6107 sub vm_commandline
{
6108 my ($storecfg, $vmid, $snapname) = @_;
6110 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6112 my ($forcemachine, $forcecpu);
6114 my $snapshot = $conf->{snapshots
}->{$snapname};
6115 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
6117 # check for machine or CPU overrides in snapshot
6118 $forcemachine = $snapshot->{runningmachine
};
6119 $forcecpu = $snapshot->{runningcpu
};
6121 $snapshot->{digest
} = $conf->{digest
}; # keep file digest for API
6126 my $defaults = load_defaults
();
6128 my $cmd = config_to_command
($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
6130 return PVE
::Tools
::cmd2string
($cmd);
6134 my ($vmid, $skiplock) = @_;
6136 PVE
::QemuConfig-
>lock_config($vmid, sub {
6138 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6140 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6142 mon_cmd
($vmid, "system_reset");
6146 sub get_vm_volumes
{
6150 foreach_volid
($conf, sub {
6151 my ($volid, $attr) = @_;
6153 return if $volid =~ m
|^/|;
6155 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
6158 push @$vollist, $volid;
6164 sub cleanup_pci_devices
{
6165 my ($vmid, $conf) = @_;
6167 foreach my $key (keys %$conf) {
6168 next if $key !~ m/^hostpci(\d+)$/;
6169 my $hostpciindex = $1;
6170 my $uuid = PVE
::SysFSTools
::generate_mdev_uuid
($vmid, $hostpciindex);
6171 my $d = parse_hostpci
($conf->{$key});
6173 # NOTE: avoid PVE::SysFSTools::pci_cleanup_mdev_device as it requires PCI ID and we
6174 # don't want to break ABI just for this two liner
6175 my $dev_sysfs_dir = "/sys/bus/mdev/devices/$uuid";
6177 # some nvidia vgpu driver versions want to clean the mdevs up themselves, and error
6178 # out when we do it first. so wait for 10 seconds and then try it
6179 if ($d->{ids
}->[0]->[0]->{vendor
} =~ m/^(0x)?10de$/) {
6183 PVE
::SysFSTools
::file_write
("$dev_sysfs_dir/remove", "1") if -e
$dev_sysfs_dir;
6186 PVE
::QemuServer
::PCI
::remove_pci_reservation
($vmid);
6189 sub vm_stop_cleanup
{
6190 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
6195 my $vollist = get_vm_volumes
($conf);
6196 PVE
::Storage
::deactivate_volumes
($storecfg, $vollist);
6198 if (my $tpmdrive = $conf->{tpmstate0
}) {
6199 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
6200 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
6202 PVE
::Storage
::unmap_volume
($storecfg, $tpm->{file
});
6207 foreach my $ext (qw(mon qmp pid vnc qga)) {
6208 unlink "/var/run/qemu-server/${vmid}.$ext";
6211 if ($conf->{ivshmem
}) {
6212 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
6213 # just delete it for now, VMs which have this already open do not
6214 # are affected, but new VMs will get a separated one. If this
6215 # becomes an issue we either add some sort of ref-counting or just
6216 # add a "don't delete on stop" flag to the ivshmem format.
6217 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name
} // $vmid);
6220 cleanup_pci_devices
($vmid, $conf);
6222 vmconfig_apply_pending
($vmid, $conf, $storecfg) if $apply_pending_changes;
6224 warn $@ if $@; # avoid errors - just warn
6227 # call only in locked context
6229 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
6231 my $pid = check_running
($vmid, $nocheck);
6236 $conf = PVE
::QemuConfig-
>load_config($vmid);
6237 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6238 if (!defined($timeout) && $shutdown && $conf->{startup
}) {
6239 my $opts = PVE
::JSONSchema
::pve_parse_startup_order
($conf->{startup
});
6240 $timeout = $opts->{down
} if $opts->{down
};
6242 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-stop');
6247 if (defined($conf) && get_qga_key
($conf, 'enabled')) {
6248 mon_cmd
($vmid, "guest-shutdown", timeout
=> $timeout);
6250 mon_cmd
($vmid, "system_powerdown");
6253 mon_cmd
($vmid, "quit");
6259 $timeout = 60 if !defined($timeout);
6262 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6267 if ($count >= $timeout) {
6269 warn "VM still running - terminating now with SIGTERM\n";
6272 die "VM quit/powerdown failed - got timeout\n";
6275 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6279 if (!check_running
($vmid, $nocheck)) {
6280 warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
6284 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
6287 die "VM quit/powerdown failed\n";
6295 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6300 if ($count >= $timeout) {
6301 warn "VM still running - terminating now with SIGKILL\n";
6306 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6309 # Note: use $nocheck to skip tests if VM configuration file exists.
6310 # We need that when migration VMs to other nodes (files already moved)
6311 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
6313 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
6315 $force = 1 if !defined($force) && !$shutdown;
6318 my $pid = check_running
($vmid, $nocheck, $migratedfrom);
6319 kill 15, $pid if $pid;
6320 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migratedfrom);
6321 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 0);
6325 PVE
::QemuConfig-
>lock_config($vmid, sub {
6326 _do_vm_stop
($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
6331 my ($vmid, $timeout) = @_;
6333 PVE
::QemuConfig-
>lock_config($vmid, sub {
6336 # only reboot if running, as qmeventd starts it again on a stop event
6337 return if !check_running
($vmid);
6339 create_reboot_request
($vmid);
6341 my $storecfg = PVE
::Storage
::config
();
6342 _do_vm_stop
($storecfg, $vmid, undef, undef, $timeout, 1);
6346 # avoid that the next normal shutdown will be confused for a reboot
6347 clear_reboot_request
($vmid);
6353 # note: if using the statestorage parameter, the caller has to check privileges
6355 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
6362 PVE
::QemuConfig-
>lock_config($vmid, sub {
6364 $conf = PVE
::QemuConfig-
>load_config($vmid);
6366 my $is_backing_up = PVE
::QemuConfig-
>has_lock($conf, 'backup');
6367 PVE
::QemuConfig-
>check_lock($conf)
6368 if !($skiplock || $is_backing_up);
6370 die "cannot suspend to disk during backup\n"
6371 if $is_backing_up && $includestate;
6373 if ($includestate) {
6374 $conf->{lock} = 'suspending';
6375 my $date = strftime
("%Y-%m-%d", localtime(time()));
6376 $storecfg = PVE
::Storage
::config
();
6377 if (!$statestorage) {
6378 $statestorage = find_vmstate_storage
($conf, $storecfg);
6379 # check permissions for the storage
6380 my $rpcenv = PVE
::RPCEnvironment
::get
();
6381 if ($rpcenv->{type
} ne 'cli') {
6382 my $authuser = $rpcenv->get_user();
6383 $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
6388 $vmstate = PVE
::QemuConfig-
>__snapshot_save_vmstate(
6389 $vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
6390 $path = PVE
::Storage
::path
($storecfg, $vmstate);
6391 PVE
::QemuConfig-
>write_config($vmid, $conf);
6393 mon_cmd
($vmid, "stop");
6397 if ($includestate) {
6399 PVE
::Storage
::activate_volumes
($storecfg, [$vmstate]);
6402 set_migration_caps
($vmid, 1);
6403 mon_cmd
($vmid, "savevm-start", statefile
=> $path);
6405 my $state = mon_cmd
($vmid, "query-savevm");
6406 if (!$state->{status
}) {
6407 die "savevm not active\n";
6408 } elsif ($state->{status
} eq 'active') {
6411 } elsif ($state->{status
} eq 'completed') {
6412 print "State saved, quitting\n";
6414 } elsif ($state->{status
} eq 'failed' && $state->{error
}) {
6415 die "query-savevm failed with error '$state->{error}'\n"
6417 die "query-savevm returned status '$state->{status}'\n";
6423 PVE
::QemuConfig-
>lock_config($vmid, sub {
6424 $conf = PVE
::QemuConfig-
>load_config($vmid);
6426 # cleanup, but leave suspending lock, to indicate something went wrong
6428 mon_cmd
($vmid, "savevm-end");
6429 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6430 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6431 delete $conf->@{qw(vmstate runningmachine runningcpu)};
6432 PVE
::QemuConfig-
>write_config($vmid, $conf);
6438 die "lock changed unexpectedly\n"
6439 if !PVE
::QemuConfig-
>has_lock($conf, 'suspending');
6441 mon_cmd
($vmid, "quit");
6442 $conf->{lock} = 'suspended';
6443 PVE
::QemuConfig-
>write_config($vmid, $conf);
6448 # $nocheck is set when called as part of a migration - in this context the
6449 # location of the config file (source or target node) is not deterministic,
6450 # since migration cannot wait for pmxcfs to process the rename
6452 my ($vmid, $skiplock, $nocheck) = @_;
6454 PVE
::QemuConfig-
>lock_config($vmid, sub {
6455 my $res = mon_cmd
($vmid, 'query-status');
6456 my $resume_cmd = 'cont';
6460 $conf = eval { PVE
::QemuConfig-
>load_config($vmid) }; # try on target node
6462 my $vmlist = PVE
::Cluster
::get_vmlist
();
6463 if (exists($vmlist->{ids
}->{$vmid})) {
6464 my $node = $vmlist->{ids
}->{$vmid}->{node
};
6465 $conf = eval { PVE
::QemuConfig-
>load_config($vmid, $node) }; # try on source node
6468 PVE
::Cluster
::cfs_update
(); # vmlist was wrong, invalidate cache
6469 $conf = PVE
::QemuConfig-
>load_config($vmid); # last try on target node again
6473 $conf = PVE
::QemuConfig-
>load_config($vmid);
6476 if ($res->{status
}) {
6477 return if $res->{status
} eq 'running'; # job done, go home
6478 $resume_cmd = 'system_wakeup' if $res->{status
} eq 'suspended';
6479 $reset = 1 if $res->{status
} eq 'shutdown';
6483 PVE
::QemuConfig-
>check_lock($conf)
6484 if !($skiplock || PVE
::QemuConfig-
>has_lock($conf, 'backup'));
6488 # required if a VM shuts down during a backup and we get a resume
6489 # request before the backup finishes for example
6490 mon_cmd
($vmid, "system_reset");
6493 add_nets_bridge_fdb
($conf, $vmid) if $resume_cmd eq 'cont';
6495 mon_cmd
($vmid, $resume_cmd);
6500 my ($vmid, $skiplock, $key) = @_;
6502 PVE
::QemuConfig-
>lock_config($vmid, sub {
6504 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6506 # there is no qmp command, so we use the human monitor command
6507 my $res = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "sendkey $key");
6508 die $res if $res ne '';
6512 sub check_bridge_access
{
6513 my ($rpcenv, $authuser, $conf) = @_;
6515 return 1 if $authuser eq 'root@pam';
6517 for my $opt (sort keys $conf->%*) {
6518 next if $opt !~ m/^net\d+$/;
6519 my $net = parse_net
($conf->{$opt});
6520 my ($bridge, $tag, $trunks) = $net->@{'bridge', 'tag', 'trunks'};
6521 PVE
::GuestHelpers
::check_vnet_access
($rpcenv, $authuser, $bridge, $tag, $trunks);
6526 sub check_mapping_access
{
6527 my ($rpcenv, $user, $conf) = @_;
6529 for my $opt (keys $conf->%*) {
6530 if ($opt =~ m/^usb\d+$/) {
6531 my $device = PVE
::JSONSchema
::parse_property_string
('pve-qm-usb', $conf->{$opt});
6532 if (my $host = $device->{host
}) {
6533 die "only root can set '$opt' config for real devices\n"
6534 if $host !~ m/^spice$/i && $user ne 'root@pam';
6535 } elsif ($device->{mapping
}) {
6536 $rpcenv->check_full($user, "/mapping/usb/$device->{mapping}", ['Mapping.Use']);
6538 die "either 'host' or 'mapping' must be set.\n";
6540 } elsif ($opt =~ m/^hostpci\d+$/) {
6541 my $device = PVE
::JSONSchema
::parse_property_string
('pve-qm-hostpci', $conf->{$opt});
6542 if ($device->{host
}) {
6543 die "only root can set '$opt' config for non-mapped devices\n" if $user ne 'root@pam';
6544 } elsif ($device->{mapping
}) {
6545 $rpcenv->check_full($user, "/mapping/pci/$device->{mapping}", ['Mapping.Use']);
6547 die "either 'host' or 'mapping' must be set.\n";
6553 sub check_restore_permissions
{
6554 my ($rpcenv, $user, $conf) = @_;
6556 check_bridge_access
($rpcenv, $user, $conf);
6557 check_mapping_access
($rpcenv, $user, $conf);
6559 # vzdump restore implementaion
6561 sub tar_archive_read_firstfile
{
6562 my $archive = shift;
6564 die "ERROR: file '$archive' does not exist\n" if ! -f
$archive;
6566 # try to detect archive type first
6567 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6568 die "unable to open file '$archive'\n";
6569 my $firstfile = <$fh>;
6573 die "ERROR: archive contaions no data\n" if !$firstfile;
6579 sub tar_restore_cleanup
{
6580 my ($storecfg, $statfile) = @_;
6582 print STDERR
"starting cleanup\n";
6584 if (my $fd = IO
::File-
>new($statfile, "r")) {
6585 while (defined(my $line = <$fd>)) {
6586 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6589 if ($volid =~ m
|^/|) {
6590 unlink $volid || die 'unlink failed\n';
6592 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6594 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6596 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6598 print STDERR
"unable to parse line in statfile - $line";
6605 sub restore_file_archive
{
6606 my ($archive, $vmid, $user, $opts) = @_;
6608 return restore_vma_archive
($archive, $vmid, $user, $opts)
6611 my $info = PVE
::Storage
::archive_info
($archive);
6612 my $format = $opts->{format
} // $info->{format
};
6613 my $comp = $info->{compression
};
6615 # try to detect archive format
6616 if ($format eq 'tar') {
6617 return restore_tar_archive
($archive, $vmid, $user, $opts);
6619 return restore_vma_archive
($archive, $vmid, $user, $opts, $comp);
6623 # hepler to remove disks that will not be used after restore
6624 my $restore_cleanup_oldconf = sub {
6625 my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
6627 my $kept_disks = {};
6629 PVE
::QemuConfig-
>foreach_volume($oldconf, sub {
6630 my ($ds, $drive) = @_;
6632 return if drive_is_cdrom
($drive, 1);
6634 my $volid = $drive->{file
};
6635 return if !$volid || $volid =~ m
|^/|;
6637 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
6638 return if !$path || !$owner || ($owner != $vmid);
6640 # Note: only delete disk we want to restore
6641 # other volumes will become unused
6642 if ($virtdev_hash->{$ds}) {
6643 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
6648 $kept_disks->{$volid} = 1;
6652 # after the restore we have no snapshots anymore
6653 for my $snapname (keys $oldconf->{snapshots
}->%*) {
6654 my $snap = $oldconf->{snapshots
}->{$snapname};
6655 if ($snap->{vmstate
}) {
6656 eval { PVE
::Storage
::vdisk_free
($storecfg, $snap->{vmstate
}); };
6662 for my $volid (keys $kept_disks->%*) {
6663 eval { PVE
::Storage
::volume_snapshot_delete
($storecfg, $volid, $snapname); };
6669 # Helper to parse vzdump backup device hints
6671 # $rpcenv: Environment, used to ckeck storage permissions
6672 # $user: User ID, to check storage permissions
6673 # $storecfg: Storage configuration
6674 # $fh: the file handle for reading the configuration
6675 # $devinfo: should contain device sizes for all backu-up'ed devices
6676 # $options: backup options (pool, default storage)
6678 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
6679 my $parse_backup_hints = sub {
6680 my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
6682 my $check_storage = sub { # assert if an image can be allocate
6683 my ($storeid, $scfg) = @_;
6684 die "Content type 'images' is not available on storage '$storeid'\n"
6685 if !$scfg->{content
}->{images
};
6686 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace'])
6687 if $user ne 'root@pam';
6690 my $virtdev_hash = {};
6691 while (defined(my $line = <$fh>)) {
6692 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6693 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6694 die "archive does not contain data for drive '$virtdev'\n"
6695 if !$devinfo->{$devname};
6697 if (defined($options->{storage
})) {
6698 $storeid = $options->{storage
} || 'local';
6699 } elsif (!$storeid) {
6702 $format = 'raw' if !$format;
6703 $devinfo->{$devname}->{devname
} = $devname;
6704 $devinfo->{$devname}->{virtdev
} = $virtdev;
6705 $devinfo->{$devname}->{format
} = $format;
6706 $devinfo->{$devname}->{storeid
} = $storeid;
6708 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6709 $check_storage->($storeid, $scfg); # permission and content type check
6711 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6712 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6714 my $drive = parse_drive
($virtdev, $2);
6716 if (drive_is_cloudinit
($drive)) {
6717 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
6718 $storeid = $options->{storage
} if defined ($options->{storage
});
6719 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6720 my $format = qemu_img_format
($scfg, $volname); # has 'raw' fallback
6722 $check_storage->($storeid, $scfg); # permission and content type check
6724 $virtdev_hash->{$virtdev} = {
6726 storeid
=> $storeid,
6727 size
=> PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
,
6734 return $virtdev_hash;
6737 # Helper to allocate and activate all volumes required for a restore
6739 # $storecfg: Storage configuration
6740 # $virtdev_hash: as returned by parse_backup_hints()
6742 # Returns: { $virtdev => $volid }
6743 my $restore_allocate_devices = sub {
6744 my ($storecfg, $virtdev_hash, $vmid) = @_;
6747 foreach my $virtdev (sort keys %$virtdev_hash) {
6748 my $d = $virtdev_hash->{$virtdev};
6749 my $alloc_size = int(($d->{size
} + 1024 - 1)/1024);
6750 my $storeid = $d->{storeid
};
6751 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6753 # test if requested format is supported
6754 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
6755 my $supported = grep { $_ eq $d->{format
} } @$validFormats;
6756 $d->{format
} = $defFormat if !$supported;
6759 if ($d->{is_cloudinit
}) {
6760 $name = "vm-$vmid-cloudinit";
6761 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6762 if ($scfg->{path
}) {
6763 $name .= ".$d->{format}";
6767 my $volid = PVE
::Storage
::vdisk_alloc
(
6768 $storecfg, $storeid, $vmid, $d->{format
}, $name, $alloc_size);
6770 print STDERR
"new volume ID is '$volid'\n";
6771 $d->{volid
} = $volid;
6773 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
6775 $map->{$virtdev} = $volid;
6781 sub restore_update_config_line
{
6782 my ($cookie, $map, $line, $unique) = @_;
6784 return '' if $line =~ m/^\#qmdump\#/;
6785 return '' if $line =~ m/^\#vzdump\#/;
6786 return '' if $line =~ m/^lock:/;
6787 return '' if $line =~ m/^unused\d+:/;
6788 return '' if $line =~ m/^parent:/;
6792 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
6793 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6794 # try to convert old 1.X settings
6795 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6796 foreach my $devconfig (PVE
::Tools
::split_list
($ethcfg)) {
6797 my ($model, $macaddr) = split(/\=/, $devconfig);
6798 $macaddr = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if !$macaddr || $unique;
6801 bridge
=> "vmbr$ind",
6802 macaddr
=> $macaddr,
6804 my $netstr = print_net
($net);
6806 $res .= "net$cookie->{netcount}: $netstr\n";
6807 $cookie->{netcount
}++;
6809 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6810 my ($id, $netstr) = ($1, $2);
6811 my $net = parse_net
($netstr);
6812 $net->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if $net->{macaddr
};
6813 $netstr = print_net
($net);
6814 $res .= "$id: $netstr\n";
6815 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk|tpmstate)\d+):\s*(\S+)\s*$/) {
6818 my $di = parse_drive
($virtdev, $value);
6819 if (defined($di->{backup
}) && !$di->{backup
}) {
6821 } elsif ($map->{$virtdev}) {
6822 delete $di->{format
}; # format can change on restore
6823 $di->{file
} = $map->{$virtdev};
6824 $value = print_drive
($di);
6825 $res .= "$virtdev: $value\n";
6829 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6831 if ($vmgenid ne '0') {
6832 # always generate a new vmgenid if there was a valid one setup
6833 $vmgenid = generate_uuid
();
6835 $res .= "vmgenid: $vmgenid\n";
6836 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6837 my ($uuid, $uuid_str);
6838 UUID
::generate
($uuid);
6839 UUID
::unparse
($uuid, $uuid_str);
6840 my $smbios1 = parse_smbios1
($2);
6841 $smbios1->{uuid
} = $uuid_str;
6842 $res .= $1.print_smbios1
($smbios1)."\n";
6850 my $restore_deactivate_volumes = sub {
6851 my ($storecfg, $virtdev_hash) = @_;
6854 for my $dev (values $virtdev_hash->%*) {
6855 push $vollist->@*, $dev->{volid
} if $dev->{volid
};
6858 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
6859 print STDERR
$@ if $@;
6862 my $restore_destroy_volumes = sub {
6863 my ($storecfg, $virtdev_hash) = @_;
6865 for my $dev (values $virtdev_hash->%*) {
6866 my $volid = $dev->{volid
} or next;
6868 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6869 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6871 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6875 sub restore_merge_config
{
6876 my ($filename, $backup_conf_raw, $override_conf) = @_;
6878 my $backup_conf = parse_vm_config
($filename, $backup_conf_raw);
6879 for my $key (keys $override_conf->%*) {
6880 $backup_conf->{$key} = $override_conf->{$key};
6883 return $backup_conf;
6887 my ($cfg, $vmid) = @_;
6889 my $info = PVE
::Storage
::vdisk_list
($cfg, undef, $vmid, undef, 'images');
6891 my $volid_hash = {};
6892 foreach my $storeid (keys %$info) {
6893 foreach my $item (@{$info->{$storeid}}) {
6894 next if !($item->{volid
} && $item->{size
});
6895 $item->{path
} = PVE
::Storage
::path
($cfg, $item->{volid
});
6896 $volid_hash->{$item->{volid
}} = $item;
6903 sub update_disk_config
{
6904 my ($vmid, $conf, $volid_hash) = @_;
6907 my $prefix = "VM $vmid";
6909 # used and unused disks
6910 my $referenced = {};
6912 # Note: it is allowed to define multiple storages with same path (alias), so
6913 # we need to check both 'volid' and real 'path' (two different volid can point
6914 # to the same path).
6916 my $referencedpath = {};
6919 PVE
::QemuConfig-
>foreach_volume($conf, sub {
6920 my ($opt, $drive) = @_;
6922 my $volid = $drive->{file
};
6924 my $volume = $volid_hash->{$volid};
6926 # mark volid as "in-use" for next step
6927 $referenced->{$volid} = 1;
6928 if ($volume && (my $path = $volume->{path
})) {
6929 $referencedpath->{$path} = 1;
6932 return if drive_is_cdrom
($drive);
6935 my ($updated, $msg) = PVE
::QemuServer
::Drive
::update_disksize
($drive, $volume->{size
});
6936 if (defined($updated)) {
6938 $conf->{$opt} = print_drive
($updated);
6939 print "$prefix ($opt): $msg\n";
6943 # remove 'unusedX' entry if volume is used
6944 PVE
::QemuConfig-
>foreach_unused_volume($conf, sub {
6945 my ($opt, $drive) = @_;
6947 my $volid = $drive->{file
};
6951 $path = $volid_hash->{$volid}->{path
} if $volid_hash->{$volid};
6952 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6953 print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
6955 delete $conf->{$opt};
6958 $referenced->{$volid} = 1;
6959 $referencedpath->{$path} = 1 if $path;
6962 foreach my $volid (sort keys %$volid_hash) {
6963 next if $volid =~ m/vm-$vmid-state-/;
6964 next if $referenced->{$volid};
6965 my $path = $volid_hash->{$volid}->{path
};
6966 next if !$path; # just to be sure
6967 next if $referencedpath->{$path};
6969 my $key = PVE
::QemuConfig-
>add_unused_volume($conf, $volid);
6970 print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
6971 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6978 my ($vmid, $nolock, $dryrun) = @_;
6980 my $cfg = PVE
::Storage
::config
();
6982 print "rescan volumes...\n";
6983 my $volid_hash = scan_volids
($cfg, $vmid);
6985 my $updatefn = sub {
6988 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6990 PVE
::QemuConfig-
>check_lock($conf);
6993 foreach my $volid (keys %$volid_hash) {
6994 my $info = $volid_hash->{$volid};
6995 $vm_volids->{$volid} = $info if $info->{vmid
} && $info->{vmid
} == $vmid;
6998 my $changes = update_disk_config
($vmid, $conf, $vm_volids);
7000 PVE
::QemuConfig-
>write_config($vmid, $conf) if $changes && !$dryrun;
7003 if (defined($vmid)) {
7007 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
7010 my $vmlist = config_list
();
7011 foreach my $vmid (keys %$vmlist) {
7015 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
7021 sub restore_proxmox_backup_archive
{
7022 my ($archive, $vmid, $user, $options) = @_;
7024 my $storecfg = PVE
::Storage
::config
();
7026 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($archive);
7027 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
7029 my $fingerprint = $scfg->{fingerprint
};
7030 my $keyfile = PVE
::Storage
::PBSPlugin
::pbs_encryption_key_file_name
($storecfg, $storeid);
7032 my $repo = PVE
::PBSClient
::get_repository
($scfg);
7033 my $namespace = $scfg->{namespace
};
7035 # This is only used for `pbs-restore` and the QEMU PBS driver (live-restore)
7036 my $password = PVE
::Storage
::PBSPlugin
::pbs_get_password
($scfg, $storeid);
7037 local $ENV{PBS_PASSWORD
} = $password;
7038 local $ENV{PBS_FINGERPRINT
} = $fingerprint if defined($fingerprint);
7040 my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
7041 PVE
::Storage
::parse_volname
($storecfg, $archive);
7043 die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
7045 die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
7047 my $tmpdir = "/var/tmp/vzdumptmp$$";
7051 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7052 # disable interrupts (always do cleanups)
7056 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7058 # Note: $oldconf is undef if VM does not exists
7059 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7060 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7061 my $new_conf_raw = '';
7063 my $rpcenv = PVE
::RPCEnvironment
::get
();
7064 my $devinfo = {}; # info about drives included in backup
7065 my $virtdev_hash = {}; # info about allocated drives
7073 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7075 my $cfgfn = "$tmpdir/qemu-server.conf";
7076 my $firewall_config_fn = "$tmpdir/fw.conf";
7077 my $index_fn = "$tmpdir/index.json";
7079 my $cmd = "restore";
7081 my $param = [$pbs_backup_name, "index.json", $index_fn];
7082 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7083 my $index = PVE
::Tools
::file_get_contents
($index_fn);
7084 $index = decode_json
($index);
7086 foreach my $info (@{$index->{files
}}) {
7087 if ($info->{filename
} =~ m/^(drive-\S+).img.fidx$/) {
7089 if ($info->{size
} =~ m/^(\d+)$/) { # untaint size
7090 $devinfo->{$devname}->{size
} = $1;
7092 die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
7097 my $is_qemu_server_backup = scalar(
7098 grep { $_->{filename
} eq 'qemu-server.conf.blob' } @{$index->{files
}}
7100 if (!$is_qemu_server_backup) {
7101 die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
7103 my $has_firewall_config = scalar(grep { $_->{filename
} eq 'fw.conf.blob' } @{$index->{files
}});
7105 $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
7106 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7108 if ($has_firewall_config) {
7109 $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
7110 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7112 my $pve_firewall_dir = '/etc/pve/firewall';
7113 mkdir $pve_firewall_dir; # make sure the dir exists
7114 PVE
::Tools
::file_copy
($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
7117 my $fh = IO
::File-
>new($cfgfn, "r") ||
7118 die "unable to read qemu-server.conf - $!\n";
7120 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
7122 # fixme: rate limit?
7124 # create empty/temp config
7125 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\nlock: create");
7127 $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
7130 my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
7132 foreach my $virtdev (sort keys %$virtdev_hash) {
7133 my $d = $virtdev_hash->{$virtdev};
7134 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7136 # this fails if storage is unavailable
7137 my $volid = $d->{volid
};
7138 my $path = PVE
::Storage
::path
($storecfg, $volid);
7140 # for live-restore we only want to preload the efidisk and TPM state
7141 next if $options->{live
} && $virtdev ne 'efidisk0' && $virtdev ne 'tpmstate0';
7144 if (defined(my $ns = $scfg->{namespace
})) {
7145 @ns_arg = ('--ns', $ns);
7148 my $pbs_restore_cmd = [
7149 '/usr/bin/pbs-restore',
7150 '--repository', $repo,
7153 "$d->{devname}.img.fidx",
7158 push @$pbs_restore_cmd, '--format', $d->{format
} if $d->{format
};
7159 push @$pbs_restore_cmd, '--keyfile', $keyfile if -e
$keyfile;
7161 if (PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $volid)) {
7162 push @$pbs_restore_cmd, '--skip-zero';
7165 my $dbg_cmdstring = PVE
::Tools
::cmd2string
($pbs_restore_cmd);
7166 print "restore proxmox backup image: $dbg_cmdstring\n";
7167 run_command
($pbs_restore_cmd);
7170 $fh->seek(0, 0) || die "seek failed - $!\n";
7172 my $cookie = { netcount
=> 0 };
7173 while (defined(my $line = <$fh>)) {
7174 $new_conf_raw .= restore_update_config_line
(
7186 if ($err || !$options->{live
}) {
7187 $restore_deactivate_volumes->($storecfg, $virtdev_hash);
7193 $restore_destroy_volumes->($storecfg, $virtdev_hash);
7197 if ($options->{live
}) {
7198 # keep lock during live-restore
7199 $new_conf_raw .= "\nlock: create";
7202 my $new_conf = restore_merge_config
($conffile, $new_conf_raw, $options->{override_conf
});
7203 check_restore_permissions
($rpcenv, $user, $new_conf);
7204 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7206 eval { rescan
($vmid, 1); };
7209 PVE
::AccessControl
::add_vm_to_pool
($vmid, $options->{pool
}) if $options->{pool
};
7211 if ($options->{live
}) {
7217 local $SIG{PIPE
} = sub { die "got signal ($!) - abort\n"; };
7219 my $conf = PVE
::QemuConfig-
>load_config($vmid);
7220 die "cannot do live-restore for template\n" if PVE
::QemuConfig-
>is_template($conf);
7222 # these special drives are already restored before start
7223 delete $devinfo->{'drive-efidisk0'};
7224 delete $devinfo->{'drive-tpmstate0-backup'};
7228 keyfile
=> $keyfile,
7229 snapshot
=> $pbs_backup_name,
7230 namespace
=> $namespace,
7232 pbs_live_restore
($vmid, $conf, $storecfg, $devinfo, $pbs_opts);
7234 PVE
::QemuConfig-
>remove_lock($vmid, "create");
7238 sub pbs_live_restore
{
7239 my ($vmid, $conf, $storecfg, $restored_disks, $opts) = @_;
7241 print "starting VM for live-restore\n";
7242 print "repository: '$opts->{repo}', snapshot: '$opts->{snapshot}'\n";
7244 my $pbs_backing = {};
7245 for my $ds (keys %$restored_disks) {
7246 $ds =~ m/^drive-(.*)$/;
7248 $pbs_backing->{$confname} = {
7249 repository
=> $opts->{repo
},
7250 snapshot
=> $opts->{snapshot
},
7251 archive
=> "$ds.img.fidx",
7253 $pbs_backing->{$confname}->{keyfile
} = $opts->{keyfile
} if -e
$opts->{keyfile
};
7254 $pbs_backing->{$confname}->{namespace
} = $opts->{namespace
} if defined($opts->{namespace
});
7256 my $drive = parse_drive
($confname, $conf->{$confname});
7257 print "restoring '$ds' to '$drive->{file}'\n";
7260 my $drives_streamed = 0;
7262 # make sure HA doesn't interrupt our restore by stopping the VM
7263 if (PVE
::HA
::Config
::vm_is_ha_managed
($vmid)) {
7264 run_command
(['ha-manager', 'set', "vm:$vmid", '--state', 'started']);
7267 # start VM with backing chain pointing to PBS backup, environment vars for PBS driver
7268 # in QEMU (PBS_PASSWORD and PBS_FINGERPRINT) are already set by our caller
7269 vm_start_nolock
($storecfg, $vmid, $conf, {paused
=> 1, 'pbs-backing' => $pbs_backing}, {});
7271 my $qmeventd_fd = register_qmeventd_handle
($vmid);
7273 # begin streaming, i.e. data copy from PBS to target disk for every vol,
7274 # this will effectively collapse the backing image chain consisting of
7275 # [target <- alloc-track -> PBS snapshot] to just [target] (alloc-track
7276 # removes itself once all backing images vanish with 'auto-remove=on')
7278 for my $ds (sort keys %$restored_disks) {
7279 my $job_id = "restore-$ds";
7280 mon_cmd
($vmid, 'block-stream',
7281 'job-id' => $job_id,
7284 $jobs->{$job_id} = {};
7287 mon_cmd
($vmid, 'cont');
7288 qemu_drive_mirror_monitor
($vmid, undef, $jobs, 'auto', 0, 'stream');
7290 print "restore-drive jobs finished successfully, removing all tracking block devices"
7291 ." to disconnect from Proxmox Backup Server\n";
7293 for my $ds (sort keys %$restored_disks) {
7294 mon_cmd
($vmid, 'blockdev-del', 'node-name' => "$ds-pbs");
7297 close($qmeventd_fd);
7303 warn "An error occurred during live-restore: $err\n";
7304 _do_vm_stop
($storecfg, $vmid, 1, 1, 10, 0, 1);
7305 die "live-restore failed\n";
7309 sub restore_vma_archive
{
7310 my ($archive, $vmid, $user, $opts, $comp) = @_;
7312 my $readfrom = $archive;
7314 my $cfg = PVE
::Storage
::config
();
7316 my $bwlimit = $opts->{bwlimit
};
7318 my $dbg_cmdstring = '';
7319 my $add_pipe = sub {
7321 push @$commands, $cmd;
7322 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
7323 $dbg_cmdstring .= PVE
::Tools
::cmd2string
($cmd);
7328 if ($archive eq '-') {
7331 # If we use a backup from a PVE defined storage we also consider that
7332 # storage's rate limit:
7333 my (undef, $volid) = PVE
::Storage
::path_to_volume_id
($cfg, $archive);
7334 if (defined($volid)) {
7335 my ($sid, undef) = PVE
::Storage
::parse_volume_id
($volid);
7336 my $readlimit = PVE
::Storage
::get_bandwidth_limit
('restore', [$sid], $bwlimit);
7338 print STDERR
"applying read rate limit: $readlimit\n";
7339 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
7340 $add_pipe->($cstream);
7346 my $info = PVE
::Storage
::decompressor_info
('vma', $comp);
7347 my $cmd = $info->{decompressor
};
7348 push @$cmd, $readfrom;
7352 my $tmpdir = "/var/tmp/vzdumptmp$$";
7355 # disable interrupts (always do cleanups)
7359 local $SIG{HUP
} = sub { warn "got interrupt - ignored\n"; };
7361 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
7362 POSIX
::mkfifo
($mapfifo, 0600);
7364 my $openfifo = sub { open($fifofh, '>', $mapfifo) or die $! };
7366 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
7371 my $devinfo = {}; # info about drives included in backup
7372 my $virtdev_hash = {}; # info about allocated drives
7374 my $rpcenv = PVE
::RPCEnvironment
::get
();
7376 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7378 # Note: $oldconf is undef if VM does not exist
7379 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7380 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7381 my $new_conf_raw = '';
7385 my $print_devmap = sub {
7386 my $cfgfn = "$tmpdir/qemu-server.conf";
7388 # we can read the config - that is already extracted
7389 my $fh = IO
::File-
>new($cfgfn, "r") ||
7390 die "unable to read qemu-server.conf - $!\n";
7392 my $fwcfgfn = "$tmpdir/qemu-server.fw";
7394 my $pve_firewall_dir = '/etc/pve/firewall';
7395 mkdir $pve_firewall_dir; # make sure the dir exists
7396 PVE
::Tools
::file_copy
($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
7399 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
7401 foreach my $info (values %{$virtdev_hash}) {
7402 my $storeid = $info->{storeid
};
7403 next if defined($storage_limits{$storeid});
7405 my $limit = PVE
::Storage
::get_bandwidth_limit
('restore', [$storeid], $bwlimit) // 0;
7406 print STDERR
"rate limit for storage $storeid: $limit KiB/s\n" if $limit;
7407 $storage_limits{$storeid} = $limit * 1024;
7410 foreach my $devname (keys %$devinfo) {
7411 die "found no device mapping information for device '$devname'\n"
7412 if !$devinfo->{$devname}->{virtdev
};
7415 # create empty/temp config
7417 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\n");
7418 $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
7422 my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
7424 # print restore information to $fifofh
7425 foreach my $virtdev (sort keys %$virtdev_hash) {
7426 my $d = $virtdev_hash->{$virtdev};
7427 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7429 my $storeid = $d->{storeid
};
7430 my $volid = $d->{volid
};
7433 if (my $limit = $storage_limits{$storeid}) {
7434 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
7437 my $write_zeros = 1;
7438 if (PVE
::Storage
::volume_has_feature
($cfg, 'sparseinit', $volid)) {
7442 my $path = PVE
::Storage
::path
($cfg, $volid);
7444 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
7446 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
7449 $fh->seek(0, 0) || die "seek failed - $!\n";
7451 my $cookie = { netcount
=> 0 };
7452 while (defined(my $line = <$fh>)) {
7453 $new_conf_raw .= restore_update_config_line
(
7470 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7471 local $SIG{ALRM
} = sub { die "got timeout\n"; };
7473 $oldtimeout = alarm($timeout);
7480 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
7481 my ($dev_id, $size, $devname) = ($1, $2, $3);
7482 $devinfo->{$devname} = { size
=> $size, dev_id
=> $dev_id };
7483 } elsif ($line =~ m/^CTIME: /) {
7484 # we correctly received the vma config, so we can disable
7485 # the timeout now for disk allocation (set to 10 minutes, so
7486 # that we always timeout if something goes wrong)
7489 print $fifofh "done\n";
7490 my $tmp = $oldtimeout || 0;
7491 $oldtimeout = undef;
7498 print "restore vma archive: $dbg_cmdstring\n";
7499 run_command
($commands, input
=> $input, outfunc
=> $parser, afterfork
=> $openfifo);
7503 alarm($oldtimeout) if $oldtimeout;
7505 $restore_deactivate_volumes->($cfg, $virtdev_hash);
7507 close($fifofh) if $fifofh;
7512 $restore_destroy_volumes->($cfg, $virtdev_hash);
7516 my $new_conf = restore_merge_config
($conffile, $new_conf_raw, $opts->{override_conf
});
7517 check_restore_permissions
($rpcenv, $user, $new_conf);
7518 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7520 eval { rescan
($vmid, 1); };
7523 PVE
::AccessControl
::add_vm_to_pool
($vmid, $opts->{pool
}) if $opts->{pool
};
7526 sub restore_tar_archive
{
7527 my ($archive, $vmid, $user, $opts) = @_;
7529 if (scalar(keys $opts->{override_conf
}->%*) > 0) {
7530 my $keystring = join(' ', keys $opts->{override_conf
}->%*);
7531 die "cannot pass along options ($keystring) when restoring from tar archive\n";
7534 if ($archive ne '-') {
7535 my $firstfile = tar_archive_read_firstfile
($archive);
7536 die "ERROR: file '$archive' does not look like a QemuServer vzdump backup\n"
7537 if $firstfile ne 'qemu-server.conf';
7540 my $storecfg = PVE
::Storage
::config
();
7542 # avoid zombie disks when restoring over an existing VM -> cleanup first
7543 # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
7544 # skiplock=1 because qmrestore has set the 'create' lock itself already
7545 my $vmcfgfn = PVE
::QemuConfig-
>config_file($vmid);
7546 destroy_vm
($storecfg, $vmid, 1, { lock => 'restore' }) if -f
$vmcfgfn;
7548 my $tocmd = "/usr/lib/qemu-server/qmextract";
7550 $tocmd .= " --storage " . PVE
::Tools
::shellquote
($opts->{storage
}) if $opts->{storage
};
7551 $tocmd .= " --pool " . PVE
::Tools
::shellquote
($opts->{pool
}) if $opts->{pool
};
7552 $tocmd .= ' --prealloc' if $opts->{prealloc
};
7553 $tocmd .= ' --info' if $opts->{info
};
7555 # tar option "xf" does not autodetect compression when read from STDIN,
7556 # so we pipe to zcat
7557 my $cmd = "zcat -f|tar xf " . PVE
::Tools
::shellquote
($archive) . " " .
7558 PVE
::Tools
::shellquote
("--to-command=$tocmd");
7560 my $tmpdir = "/var/tmp/vzdumptmp$$";
7563 local $ENV{VZDUMP_TMPDIR
} = $tmpdir;
7564 local $ENV{VZDUMP_VMID
} = $vmid;
7565 local $ENV{VZDUMP_USER
} = $user;
7567 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7568 my $new_conf_raw = '';
7570 # disable interrupts (always do cleanups)
7574 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7582 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7584 if ($archive eq '-') {
7585 print "extracting archive from STDIN\n";
7586 run_command
($cmd, input
=> "<&STDIN");
7588 print "extracting archive '$archive'\n";
7592 return if $opts->{info
};
7596 my $statfile = "$tmpdir/qmrestore.stat";
7597 if (my $fd = IO
::File-
>new($statfile, "r")) {
7598 while (defined (my $line = <$fd>)) {
7599 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
7600 $map->{$1} = $2 if $1;
7602 print STDERR
"unable to parse line in statfile - $line\n";
7608 my $confsrc = "$tmpdir/qemu-server.conf";
7610 my $srcfd = IO
::File-
>new($confsrc, "r") || die "unable to open file '$confsrc'\n";
7612 my $cookie = { netcount
=> 0 };
7613 while (defined (my $line = <$srcfd>)) {
7614 $new_conf_raw .= restore_update_config_line
(
7625 tar_restore_cleanup
($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info
};
7631 PVE
::Tools
::file_set_contents
($conffile, $new_conf_raw);
7633 PVE
::Cluster
::cfs_update
(); # make sure we read new file
7635 eval { rescan
($vmid, 1); };
7639 sub foreach_storage_used_by_vm
{
7640 my ($conf, $func) = @_;
7644 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7645 my ($ds, $drive) = @_;
7646 return if drive_is_cdrom
($drive);
7648 my $volid = $drive->{file
};
7650 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
7651 $sidhash->{$sid} = $sid if $sid;
7654 foreach my $sid (sort keys %$sidhash) {
7659 my $qemu_snap_storage = {
7662 sub do_snapshots_with_qemu
{
7663 my ($storecfg, $volid, $deviceid) = @_;
7665 return if $deviceid =~ m/tpmstate0/;
7667 my $storage_name = PVE
::Storage
::parse_volume_id
($volid);
7668 my $scfg = $storecfg->{ids
}->{$storage_name};
7669 die "could not find storage '$storage_name'\n" if !defined($scfg);
7671 if ($qemu_snap_storage->{$scfg->{type
}} && !$scfg->{krbd
}){
7675 if ($volid =~ m/\.(qcow2|qed)$/){
7682 sub qga_check_running
{
7683 my ($vmid, $nowarn) = @_;
7685 eval { mon_cmd
($vmid, "guest-ping", timeout
=> 3); };
7687 warn "QEMU Guest Agent is not running - $@" if !$nowarn;
7693 sub template_create
{
7694 my ($vmid, $conf, $disk) = @_;
7696 my $storecfg = PVE
::Storage
::config
();
7698 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7699 my ($ds, $drive) = @_;
7701 return if drive_is_cdrom
($drive);
7702 return if $disk && $ds ne $disk;
7704 my $volid = $drive->{file
};
7705 return if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
7707 my $voliddst = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
7708 $drive->{file
} = $voliddst;
7709 $conf->{$ds} = print_drive
($drive);
7710 PVE
::QemuConfig-
>write_config($vmid, $conf);
7714 sub convert_iscsi_path
{
7717 if ($path =~ m
|^iscsi
://([^/]+)/([^/]+)/(.+)$|) {
7722 my $initiator_name = get_initiator_name
();
7724 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
7725 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
7728 die "cannot convert iscsi path '$path', unkown format\n";
7731 sub qemu_img_convert
{
7732 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized, $bwlimit) = @_;
7734 my $storecfg = PVE
::Storage
::config
();
7735 my ($src_storeid, $src_volname) = PVE
::Storage
::parse_volume_id
($src_volid, 1);
7736 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid, 1);
7738 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
7742 my $src_is_iscsi = 0;
7746 PVE
::Storage
::activate_volumes
($storecfg, [$src_volid], $snapname);
7747 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7748 $src_format = qemu_img_format
($src_scfg, $src_volname);
7749 $src_path = PVE
::Storage
::path
($storecfg, $src_volid, $snapname);
7750 $src_is_iscsi = ($src_path =~ m
|^iscsi
://|);
7751 $cachemode = 'none' if $src_scfg->{type
} eq 'zfspool';
7752 } elsif (-f
$src_volid || -b
$src_volid) {
7753 $src_path = $src_volid;
7754 if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7759 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
7761 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7762 my $dst_format = qemu_img_format
($dst_scfg, $dst_volname);
7763 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7764 my $dst_is_iscsi = ($dst_path =~ m
|^iscsi
://|);
7767 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
7768 push @$cmd, '-l', "snapshot.name=$snapname"
7769 if $snapname && $src_format && $src_format eq "qcow2";
7770 push @$cmd, '-t', 'none' if $dst_scfg->{type
} eq 'zfspool';
7771 push @$cmd, '-T', $cachemode if defined($cachemode);
7772 push @$cmd, '-r', "${bwlimit}K" if defined($bwlimit);
7774 if ($src_is_iscsi) {
7775 push @$cmd, '--image-opts';
7776 $src_path = convert_iscsi_path
($src_path);
7777 } elsif ($src_format) {
7778 push @$cmd, '-f', $src_format;
7781 if ($dst_is_iscsi) {
7782 push @$cmd, '--target-image-opts';
7783 $dst_path = convert_iscsi_path
($dst_path);
7785 push @$cmd, '-O', $dst_format;
7788 push @$cmd, $src_path;
7790 if (!$dst_is_iscsi && $is_zero_initialized) {
7791 push @$cmd, "zeroinit:$dst_path";
7793 push @$cmd, $dst_path;
7798 if($line =~ m/\((\S+)\/100\
%\)/){
7800 my $transferred = int($size * $percent / 100);
7801 my $total_h = render_bytes
($size, 1);
7802 my $transferred_h = render_bytes
($transferred, 1);
7804 print "transferred $transferred_h of $total_h ($percent%)\n";
7809 eval { run_command
($cmd, timeout
=> undef, outfunc
=> $parser); };
7811 die "copy failed: $err" if $err;
7814 sub qemu_img_format
{
7815 my ($scfg, $volname) = @_;
7817 if ($scfg->{path
} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7824 sub qemu_drive_mirror
{
7825 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
7827 $jobs = {} if !$jobs;
7831 $jobs->{"drive-$drive"} = {};
7833 if ($dst_volid =~ /^nbd:/) {
7834 $qemu_target = $dst_volid;
7837 my $storecfg = PVE
::Storage
::config
();
7838 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid);
7840 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7842 $format = qemu_img_format
($dst_scfg, $dst_volname);
7844 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7846 $qemu_target = $is_zero_initialized ?
"zeroinit:$dst_path" : $dst_path;
7849 my $opts = { timeout
=> 10, device
=> "drive-$drive", mode
=> "existing", sync
=> "full", target
=> $qemu_target };
7850 $opts->{format
} = $format if $format;
7852 if (defined($src_bitmap)) {
7853 $opts->{sync
} = 'incremental';
7854 $opts->{bitmap
} = $src_bitmap;
7855 print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
7858 if (defined($bwlimit)) {
7859 $opts->{speed
} = $bwlimit * 1024;
7860 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7862 print "drive mirror is starting for drive-$drive\n";
7865 # if a job already runs for this device we get an error, catch it for cleanup
7866 eval { mon_cmd
($vmid, "drive-mirror", %$opts); };
7868 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7870 die "mirroring error: $err\n";
7873 qemu_drive_mirror_monitor
($vmid, $vmiddst, $jobs, $completion, $qga);
7876 # $completion can be either
7877 # 'complete': wait until all jobs are ready, block-job-complete them (default)
7878 # 'cancel': wait until all jobs are ready, block-job-cancel them
7879 # 'skip': wait until all jobs are ready, return with block jobs in ready state
7880 # 'auto': wait until all jobs disappear, only use for jobs which complete automatically
7881 sub qemu_drive_mirror_monitor
{
7882 my ($vmid, $vmiddst, $jobs, $completion, $qga, $op) = @_;
7884 $completion //= 'complete';
7888 my $err_complete = 0;
7890 my $starttime = time ();
7892 die "block job ('$op') timed out\n" if $err_complete > 300;
7894 my $stats = mon_cmd
($vmid, "query-block-jobs");
7897 my $running_jobs = {};
7898 for my $stat (@$stats) {
7899 next if $stat->{type
} ne $op;
7900 $running_jobs->{$stat->{device
}} = $stat;
7903 my $readycounter = 0;
7905 for my $job_id (sort keys %$jobs) {
7906 my $job = $running_jobs->{$job_id};
7908 my $vanished = !defined($job);
7909 my $complete = defined($jobs->{$job_id}->{complete
}) && $vanished;
7910 if($complete || ($vanished && $completion eq 'auto')) {
7911 print "$job_id: $op-job finished\n";
7912 delete $jobs->{$job_id};
7916 die "$job_id: '$op' has been cancelled\n" if !defined($job);
7918 my $busy = $job->{busy
};
7919 my $ready = $job->{ready
};
7920 if (my $total = $job->{len
}) {
7921 my $transferred = $job->{offset
} || 0;
7922 my $remaining = $total - $transferred;
7923 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7925 my $duration = $ctime - $starttime;
7926 my $total_h = render_bytes
($total, 1);
7927 my $transferred_h = render_bytes
($transferred, 1);
7929 my $status = sprintf(
7930 "transferred $transferred_h of $total_h ($percent%%) in %s",
7931 render_duration
($duration),
7936 $status .= ", still busy"; # shouldn't even happen? but mirror is weird
7938 $status .= ", ready";
7941 print "$job_id: $status\n" if !$jobs->{$job_id}->{ready
};
7942 $jobs->{$job_id}->{ready
} = $ready;
7945 $readycounter++ if $job->{ready
};
7948 last if scalar(keys %$jobs) == 0;
7950 if ($readycounter == scalar(keys %$jobs)) {
7951 print "all '$op' jobs are ready\n";
7953 # do the complete later (or has already been done)
7954 last if $completion eq 'skip' || $completion eq 'auto';
7956 if ($vmiddst && $vmiddst != $vmid) {
7957 my $agent_running = $qga && qga_check_running
($vmid);
7958 if ($agent_running) {
7959 print "freeze filesystem\n";
7960 eval { mon_cmd
($vmid, "guest-fsfreeze-freeze"); };
7963 print "suspend vm\n";
7964 eval { PVE
::QemuServer
::vm_suspend
($vmid, 1); };
7968 # if we clone a disk for a new target vm, we don't switch the disk
7969 PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs);
7971 if ($agent_running) {
7972 print "unfreeze filesystem\n";
7973 eval { mon_cmd
($vmid, "guest-fsfreeze-thaw"); };
7976 print "resume vm\n";
7977 eval { PVE
::QemuServer
::vm_resume
($vmid, 1, 1); };
7984 for my $job_id (sort keys %$jobs) {
7985 # try to switch the disk if source and destination are on the same guest
7986 print "$job_id: Completing block job_id...\n";
7989 if ($completion eq 'complete') {
7990 $op = 'block-job-complete';
7991 } elsif ($completion eq 'cancel') {
7992 $op = 'block-job-cancel';
7994 die "invalid completion value: $completion\n";
7996 eval { mon_cmd
($vmid, $op, device
=> $job_id) };
7997 if ($@ =~ m/cannot be completed/) {
7998 print "$job_id: block job cannot be completed, trying again.\n";
8001 print "$job_id: Completed successfully.\n";
8002 $jobs->{$job_id}->{complete
} = 1;
8013 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
8014 die "block job ($op) error: $err";
8018 sub qemu_blockjobs_cancel
{
8019 my ($vmid, $jobs) = @_;
8021 foreach my $job (keys %$jobs) {
8022 print "$job: Cancelling block job\n";
8023 eval { mon_cmd
($vmid, "block-job-cancel", device
=> $job); };
8024 $jobs->{$job}->{cancel
} = 1;
8028 my $stats = mon_cmd
($vmid, "query-block-jobs");
8030 my $running_jobs = {};
8031 foreach my $stat (@$stats) {
8032 $running_jobs->{$stat->{device
}} = $stat;
8035 foreach my $job (keys %$jobs) {
8037 if (defined($jobs->{$job}->{cancel
}) && !defined($running_jobs->{$job})) {
8038 print "$job: Done.\n";
8039 delete $jobs->{$job};
8043 last if scalar(keys %$jobs) == 0;
8049 # Check for bug #4525: drive-mirror will open the target drive with the same aio setting as the
8050 # source, but some storages have problems with io_uring, sometimes even leading to crashes.
8051 my sub clone_disk_check_io_uring
{
8052 my ($src_drive, $storecfg, $src_storeid, $dst_storeid, $use_drive_mirror) = @_;
8054 return if !$use_drive_mirror;
8056 # Don't complain when not changing storage.
8057 # Assume if it works for the source, it'll work for the target too.
8058 return if $src_storeid eq $dst_storeid;
8060 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
8061 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
8063 my $cache_direct = drive_uses_cache_direct
($src_drive);
8065 my $src_uses_io_uring;
8066 if ($src_drive->{aio
}) {
8067 $src_uses_io_uring = $src_drive->{aio
} eq 'io_uring';
8069 $src_uses_io_uring = storage_allows_io_uring_default
($src_scfg, $cache_direct);
8072 die "target storage is known to cause issues with aio=io_uring (used by current drive)\n"
8073 if $src_uses_io_uring && !storage_allows_io_uring_default
($dst_scfg, $cache_direct);
8077 my ($storecfg, $source, $dest, $full, $newvollist, $jobs, $completion, $qga, $bwlimit) = @_;
8079 my ($vmid, $running) = $source->@{qw(vmid running)};
8080 my ($src_drivename, $drive, $snapname) = $source->@{qw(drivename drive snapname)};
8082 my ($newvmid, $dst_drivename, $efisize) = $dest->@{qw(vmid drivename efisize)};
8083 my ($storage, $format) = $dest->@{qw(storage format)};
8085 my $use_drive_mirror = $full && $running && $src_drivename && !$snapname;
8087 if ($src_drivename && $dst_drivename && $src_drivename ne $dst_drivename) {
8088 die "cloning from/to EFI disk requires EFI disk\n"
8089 if $src_drivename eq 'efidisk0' || $dst_drivename eq 'efidisk0';
8090 die "cloning from/to TPM state requires TPM state\n"
8091 if $src_drivename eq 'tpmstate0' || $dst_drivename eq 'tpmstate0';
8093 # This would lead to two device nodes in QEMU pointing to the same backing image!
8094 die "cannot change drive name when cloning disk from/to the same VM\n"
8095 if $use_drive_mirror && $vmid == $newvmid;
8098 die "cannot move TPM state while VM is running\n"
8099 if $use_drive_mirror && $src_drivename eq 'tpmstate0';
8103 print "create " . ($full ?
'full' : 'linked') . " clone of drive ";
8104 print "$src_drivename " if $src_drivename;
8105 print "($drive->{file})\n";
8108 $newvolid = PVE
::Storage
::vdisk_clone
($storecfg, $drive->{file
}, $newvmid, $snapname);
8109 push @$newvollist, $newvolid;
8111 my ($src_storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
8112 my $storeid = $storage || $src_storeid;
8114 my $dst_format = resolve_dst_disk_format
($storecfg, $storeid, $volname, $format);
8118 if (drive_is_cloudinit
($drive)) {
8119 $name = "vm-$newvmid-cloudinit";
8120 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8121 if ($scfg->{path
}) {
8122 $name .= ".$dst_format";
8125 $size = PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
;
8126 } elsif ($dst_drivename eq 'efidisk0') {
8127 $size = $efisize or die "internal error - need to specify EFI disk size\n";
8128 } elsif ($dst_drivename eq 'tpmstate0') {
8129 $dst_format = 'raw';
8130 $size = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8132 clone_disk_check_io_uring
($drive, $storecfg, $src_storeid, $storeid, $use_drive_mirror);
8134 $size = PVE
::Storage
::volume_size_info
($storecfg, $drive->{file
}, 10);
8136 $newvolid = PVE
::Storage
::vdisk_alloc
(
8137 $storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024)
8139 push @$newvollist, $newvolid;
8141 PVE
::Storage
::activate_volumes
($storecfg, [$newvolid]);
8143 if (drive_is_cloudinit
($drive)) {
8144 # when cloning multiple disks (e.g. during clone_vm) it might be the last disk
8145 # if this is the case, we have to complete any block-jobs still there from
8146 # previous drive-mirrors
8147 if (($completion eq 'complete') && (scalar(keys %$jobs) > 0)) {
8148 qemu_drive_mirror_monitor
($vmid, $newvmid, $jobs, $completion, $qga);
8153 my $sparseinit = PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $newvolid);
8154 if ($use_drive_mirror) {
8155 qemu_drive_mirror
($vmid, $src_drivename, $newvolid, $newvmid, $sparseinit, $jobs,
8156 $completion, $qga, $bwlimit);
8158 if ($dst_drivename eq 'efidisk0') {
8159 # the relevant data on the efidisk may be smaller than the source
8160 # e.g. on RBD/ZFS, so we use dd to copy only the amount
8161 # that is given by the OVMF_VARS.fd
8162 my $src_path = PVE
::Storage
::path
($storecfg, $drive->{file
}, $snapname);
8163 my $dst_path = PVE
::Storage
::path
($storecfg, $newvolid);
8165 my $src_format = (PVE
::Storage
::parse_volname
($storecfg, $drive->{file
}))[6];
8167 # better for Ceph if block size is not too small, see bug #3324
8170 my $cmd = ['qemu-img', 'dd', '-n', '-O', $dst_format];
8172 if ($src_format eq 'qcow2' && $snapname) {
8173 die "cannot clone qcow2 EFI disk snapshot - requires QEMU >= 6.2\n"
8174 if !min_version
(kvm_user_version
(), 6, 2);
8175 push $cmd->@*, '-l', $snapname;
8177 push $cmd->@*, "bs=$bs", "osize=$size", "if=$src_path", "of=$dst_path";
8180 qemu_img_convert
($drive->{file
}, $newvolid, $size, $snapname, $sparseinit, $bwlimit);
8186 my $size = eval { PVE
::Storage
::volume_size_info
($storecfg, $newvolid, 10) };
8188 my $disk = dclone
($drive);
8189 delete $disk->{format
};
8190 $disk->{file
} = $newvolid;
8191 $disk->{size
} = $size if defined($size);
8196 sub get_running_qemu_version
{
8198 my $res = mon_cmd
($vmid, "query-version");
8199 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
8202 sub qemu_use_old_bios_files
{
8203 my ($machine_type) = @_;
8205 return if !$machine_type;
8207 my $use_old_bios_files = undef;
8209 if ($machine_type =~ m/^(\S+)\.pxe$/) {
8211 $use_old_bios_files = 1;
8213 my $version = extract_version
($machine_type, kvm_user_version
());
8214 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
8215 # load new efi bios files on migration. So this hack is required to allow
8216 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
8217 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
8218 $use_old_bios_files = !min_version
($version, 2, 4);
8221 return ($use_old_bios_files, $machine_type);
8224 sub get_efivars_size
{
8225 my ($conf, $efidisk) = @_;
8227 my $arch = get_vm_arch
($conf);
8228 $efidisk //= $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
8229 my $smm = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
8230 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8231 return -s
$ovmf_vars;
8234 sub update_efidisk_size
{
8237 return if !defined($conf->{efidisk0
});
8239 my $disk = PVE
::QemuServer
::parse_drive
('efidisk0', $conf->{efidisk0
});
8240 $disk->{size
} = get_efivars_size
($conf);
8241 $conf->{efidisk0
} = print_drive
($disk);
8246 sub update_tpmstate_size
{
8249 my $disk = PVE
::QemuServer
::parse_drive
('tpmstate0', $conf->{tpmstate0
});
8250 $disk->{size
} = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8251 $conf->{tpmstate0
} = print_drive
($disk);
8254 sub create_efidisk
($$$$$$$) {
8255 my ($storecfg, $storeid, $vmid, $fmt, $arch, $efidisk, $smm) = @_;
8257 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8259 my $vars_size_b = -s
$ovmf_vars;
8260 my $vars_size = PVE
::Tools
::convert_size
($vars_size_b, 'b' => 'kb');
8261 my $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
8262 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
8264 qemu_img_convert
($ovmf_vars, $volid, $vars_size_b, undef, 0);
8265 my $size = PVE
::Storage
::volume_size_info
($storecfg, $volid, 3);
8267 return ($volid, $size/1024);
8270 sub vm_iothreads_list
{
8273 my $res = mon_cmd
($vmid, 'query-iothreads');
8276 foreach my $iothread (@$res) {
8277 $iothreads->{ $iothread->{id
} } = $iothread->{"thread-id"};
8284 my ($conf, $drive) = @_;
8288 if (!$conf->{scsihw
} || ($conf->{scsihw
} =~ m/^lsi/)) {
8290 } elsif ($conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single')) {
8296 my $controller = int($drive->{index} / $maxdev);
8297 my $controller_prefix = ($conf->{scsihw
} && $conf->{scsihw
} eq 'virtio-scsi-single')
8301 return ($maxdev, $controller, $controller_prefix);
8304 sub resolve_dst_disk_format
{
8305 my ($storecfg, $storeid, $src_volname, $format) = @_;
8306 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
8309 # if no target format is specified, use the source disk format as hint
8311 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8312 $format = qemu_img_format
($scfg, $src_volname);
8318 # test if requested format is supported - else use default
8319 my $supported = grep { $_ eq $format } @$validFormats;
8320 $format = $defFormat if !$supported;
8324 # NOTE: if this logic changes, please update docs & possibly gui logic
8325 sub find_vmstate_storage
{
8326 my ($conf, $storecfg) = @_;
8328 # first, return storage from conf if set
8329 return $conf->{vmstatestorage
} if $conf->{vmstatestorage
};
8331 my ($target, $shared, $local);
8333 foreach_storage_used_by_vm
($conf, sub {
8335 my $scfg = PVE
::Storage
::storage_config
($storecfg, $sid);
8336 my $dst = $scfg->{shared
} ? \
$shared : \
$local;
8337 $$dst = $sid if !$$dst || $scfg->{path
}; # prefer file based storage
8340 # second, use shared storage where VM has at least one disk
8341 # third, use local storage where VM has at least one disk
8342 # fall back to local storage
8343 $target = $shared // $local // 'local';
8349 my ($uuid, $uuid_str);
8350 UUID
::generate
($uuid);
8351 UUID
::unparse
($uuid, $uuid_str);
8355 sub generate_smbios1_uuid
{
8356 return "uuid=".generate_uuid
();
8362 mon_cmd
($vmid, 'nbd-server-stop');
8365 sub create_reboot_request
{
8367 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
8368 or die "failed to create reboot trigger file: $!\n";
8372 sub clear_reboot_request
{
8374 my $path = "/run/qemu-server/$vmid.reboot";
8377 $res = unlink($path);
8378 die "could not remove reboot request for $vmid: $!"
8379 if !$res && $! != POSIX
::ENOENT
;
8384 sub bootorder_from_legacy
{
8385 my ($conf, $bootcfg) = @_;
8387 my $boot = $bootcfg->{legacy
} || $boot_fmt->{legacy
}->{default};
8388 my $bootindex_hash = {};
8390 foreach my $o (split(//, $boot)) {
8391 $bootindex_hash->{$o} = $i*100;
8397 PVE
::QemuConfig-
>foreach_volume($conf, sub {
8398 my ($ds, $drive) = @_;
8400 if (drive_is_cdrom
($drive, 1)) {
8401 if ($bootindex_hash->{d
}) {
8402 $bootorder->{$ds} = $bootindex_hash->{d
};
8403 $bootindex_hash->{d
} += 1;
8405 } elsif ($bootindex_hash->{c
}) {
8406 $bootorder->{$ds} = $bootindex_hash->{c
}
8407 if $conf->{bootdisk
} && $conf->{bootdisk
} eq $ds;
8408 $bootindex_hash->{c
} += 1;
8412 if ($bootindex_hash->{n
}) {
8413 for (my $i = 0; $i < $MAX_NETS; $i++) {
8414 my $netname = "net$i";
8415 next if !$conf->{$netname};
8416 $bootorder->{$netname} = $bootindex_hash->{n
};
8417 $bootindex_hash->{n
} += 1;
8424 # Generate default device list for 'boot: order=' property. Matches legacy
8425 # default boot order, but with explicit device names. This is important, since
8426 # the fallback for when neither 'order' nor the old format is specified relies
8427 # on 'bootorder_from_legacy' above, and it would be confusing if this diverges.
8428 sub get_default_bootdevices
{
8434 my $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 0);
8435 push @ret, $first if $first;
8438 $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 1);
8439 push @ret, $first if $first;
8442 for (my $i = 0; $i < $MAX_NETS; $i++) {
8443 my $netname = "net$i";
8444 next if !$conf->{$netname};
8445 push @ret, $netname;
8452 sub device_bootorder
{
8455 return bootorder_from_legacy
($conf) if !defined($conf->{boot
});
8457 my $boot = parse_property_string
($boot_fmt, $conf->{boot
});
8460 if (!defined($boot) || $boot->{legacy
}) {
8461 $bootorder = bootorder_from_legacy
($conf, $boot);
8462 } elsif ($boot->{order
}) {
8463 my $i = 100; # start at 100 to allow user to insert devices before us with -args
8464 for my $dev (PVE
::Tools
::split_list
($boot->{order
})) {
8465 $bootorder->{$dev} = $i++;
8472 sub register_qmeventd_handle
{
8476 my $peer = "/var/run/qmeventd.sock";
8481 $fh = IO
::Socket
::UNIX-
>new(Peer
=> $peer, Blocking
=> 0, Timeout
=> 1);
8483 if ($! != EINTR
&& $! != EAGAIN
) {
8484 die "unable to connect to qmeventd socket (vmid: $vmid) - $!\n";
8487 die "unable to connect to qmeventd socket (vmid: $vmid) - timeout "
8488 . "after $count retries\n";
8493 # send handshake to mark VM as backing up
8494 print $fh to_json
({vzdump
=> {vmid
=> "$vmid"}});
8496 # return handle to be closed later when inhibit is no longer required
8500 # bash completion helper
8502 sub complete_backup_archives
{
8503 my ($cmdname, $pname, $cvalue) = @_;
8505 my $cfg = PVE
::Storage
::config
();
8509 if ($cvalue =~ m/^([^:]+):/) {
8513 my $data = PVE
::Storage
::template_list
($cfg, $storeid, 'backup');
8516 foreach my $id (keys %$data) {
8517 foreach my $item (@{$data->{$id}}) {
8518 next if $item->{format
} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
8519 push @$res, $item->{volid
} if defined($item->{volid
});
8526 my $complete_vmid_full = sub {
8529 my $idlist = vmstatus
();
8533 foreach my $id (keys %$idlist) {
8534 my $d = $idlist->{$id};
8535 if (defined($running)) {
8536 next if $d->{template
};
8537 next if $running && $d->{status
} ne 'running';
8538 next if !$running && $d->{status
} eq 'running';
8547 return &$complete_vmid_full();
8550 sub complete_vmid_stopped
{
8551 return &$complete_vmid_full(0);
8554 sub complete_vmid_running
{
8555 return &$complete_vmid_full(1);
8558 sub complete_storage
{
8560 my $cfg = PVE
::Storage
::config
();
8561 my $ids = $cfg->{ids
};
8564 foreach my $sid (keys %$ids) {
8565 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, undef, 1);
8566 next if !$ids->{$sid}->{content
}->{images
};
8573 sub complete_migration_storage
{
8574 my ($cmd, $param, $current_value, $all_args) = @_;
8576 my $targetnode = @$all_args[1];
8578 my $cfg = PVE
::Storage
::config
();
8579 my $ids = $cfg->{ids
};
8582 foreach my $sid (keys %$ids) {
8583 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, $targetnode, 1);
8584 next if !$ids->{$sid}->{content
}->{images
};
8593 my $qmpstatus = eval {
8594 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid);
8595 mon_cmd
($vmid, "query-status");
8598 return $qmpstatus && $qmpstatus->{status
} eq "paused";
8601 sub check_volume_storage_type
{
8602 my ($storecfg, $vol) = @_;
8604 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($vol);
8605 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8606 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $vol);
8608 die "storage '$storeid' does not support content-type '$vtype'\n"
8609 if !$scfg->{content
}->{$vtype};
8614 sub add_nets_bridge_fdb
{
8615 my ($conf, $vmid) = @_;
8617 for my $opt (keys %$conf) {
8618 next if $opt !~ m/^net(\d+)$/;
8619 my $iface = "tap${vmid}i$1";
8620 # NOTE: expect setups with learning off to *not* use auto-random-generation of MAC on start
8621 my $net = parse_net
($conf->{$opt}, 1) or next;
8623 my $mac = $net->{macaddr
};
8625 log_warn
("MAC learning disabled, but vNIC '$iface' has no static MAC to add to forwarding DB!")
8626 if !file_read_firstline
("/sys/class/net/$iface/brport/learning");
8630 my $bridge = $net->{bridge
};
8632 log_warn
("Interface '$iface' not attached to any bridge.");
8636 PVE
::Network
::SDN
::Zones
::add_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8637 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8638 PVE
::Network
::add_bridge_fdb
($iface, $mac, $net->{firewall
});
8643 sub del_nets_bridge_fdb
{
8644 my ($conf, $vmid) = @_;
8646 for my $opt (keys %$conf) {
8647 next if $opt !~ m/^net(\d+)$/;
8648 my $iface = "tap${vmid}i$1";
8650 my $net = parse_net
($conf->{$opt}) or next;
8651 my $mac = $net->{macaddr
} or next;
8653 my $bridge = $net->{bridge
};
8655 PVE
::Network
::SDN
::Zones
::del_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8656 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8657 PVE
::Network
::del_bridge_fdb
($iface, $mac, $net->{firewall
});