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);
38 use PVE
::JSONSchema
qw(get_standard_option parse_property_string);
41 use PVE
::RESTEnvironment
qw(log_warn);
42 use PVE
::RPCEnvironment
;
46 use PVE
::Tools
qw(run_command file_read_firstline file_get_contents dir_glob_foreach get_host_arch $IPV6RE);
50 use PVE
::QemuServer
::Helpers
qw(min_version config_aware_timeout windows_version);
51 use PVE
::QemuServer
::Cloudinit
;
52 use PVE
::QemuServer
::CGroup
;
53 use PVE
::QemuServer
::CPUConfig
qw(print_cpu_device get_cpu_options);
54 use PVE
::QemuServer
::Drive
qw(is_valid_drivename drive_is_cloudinit drive_is_cdrom drive_is_read_only parse_drive print_drive);
55 use PVE
::QemuServer
::Machine
;
56 use PVE
::QemuServer
::Memory
;
57 use PVE
::QemuServer
::Monitor
qw(mon_cmd);
58 use PVE
::QemuServer
::PCI
qw(print_pci_addr print_pcie_addr print_pcie_root_port parse_hostpci);
59 use PVE
::QemuServer
::USB
qw(parse_usb_device);
63 require PVE
::Network
::SDN
::Zones
;
67 my $EDK2_FW_BASE = '/usr/share/pve-edk2-firmware/';
71 "$EDK2_FW_BASE/OVMF_CODE_4M.fd",
72 "$EDK2_FW_BASE/OVMF_VARS_4M.fd",
75 "$EDK2_FW_BASE/OVMF_CODE_4M.fd",
76 "$EDK2_FW_BASE/OVMF_VARS_4M.ms.fd",
79 "$EDK2_FW_BASE/OVMF_CODE_4M.secboot.fd",
80 "$EDK2_FW_BASE/OVMF_VARS_4M.fd",
83 "$EDK2_FW_BASE/OVMF_CODE_4M.secboot.fd",
84 "$EDK2_FW_BASE/OVMF_VARS_4M.ms.fd",
87 "$EDK2_FW_BASE/OVMF_CODE.fd",
88 "$EDK2_FW_BASE/OVMF_VARS.fd",
93 "$EDK2_FW_BASE/AAVMF_CODE.fd",
94 "$EDK2_FW_BASE/AAVMF_VARS.fd",
99 my $cpuinfo = PVE
::ProcFSTools
::read_cpuinfo
();
101 # Note about locking: we use flock on the config file protect against concurent actions.
102 # Aditionaly, we have a 'lock' setting in the config file. This can be set to 'migrate',
103 # 'backup', 'snapshot' or 'rollback'. Most actions are not allowed when such lock is set.
104 # But you can ignore this kind of lock with the --skiplock flag.
112 PVE
::JSONSchema
::register_standard_option
('pve-qm-stateuri', {
113 description
=> "Some command save/restore state from this location.",
119 PVE
::JSONSchema
::register_standard_option
('pve-qemu-machine', {
120 description
=> "Specifies the QEMU machine type.",
122 pattern
=> '(pc|pc(-i440fx)?-\d+(\.\d+)+(\+pve\d+)?(\.pxe)?|q35|pc-q35-\d+(\.\d+)+(\+pve\d+)?(\.pxe)?|virt(?:-\d+(\.\d+)+)?(\+pve\d+)?)',
127 # FIXME: remove in favor of just using the INotify one, it's cached there exactly the same way
130 $nodename_cache //= PVE
::INotify
::nodename
();
131 return $nodename_cache;
138 enum
=> [qw(i6300esb ib700)],
139 description
=> "Watchdog type to emulate.",
140 default => 'i6300esb',
145 enum
=> [qw(reset shutdown poweroff pause debug none)],
146 description
=> "The action to perform if after activation the guest fails to poll the watchdog in time.",
150 PVE
::JSONSchema
::register_format
('pve-qm-watchdog', $watchdog_fmt);
154 description
=> "Enable/disable communication with a QEMU Guest Agent (QGA) running in the VM.",
159 fstrim_cloned_disks
=> {
160 description
=> "Run fstrim after moving a disk or migrating the VM.",
165 'freeze-fs-on-backup' => {
166 description
=> "Freeze/thaw guest filesystems on backup for consistency.",
172 description
=> "Select the agent type",
176 enum
=> [qw(virtio isa)],
182 description
=> "Select the VGA type.",
187 enum
=> [qw(cirrus qxl qxl2 qxl3 qxl4 none serial0 serial1 serial2 serial3 std virtio virtio-gl vmware)],
190 description
=> "Sets the VGA memory (in MiB). Has no effect with serial display.",
202 description
=> "The size of the file in MB.",
206 pattern
=> '[a-zA-Z0-9\-]+',
208 format_description
=> 'string',
209 description
=> "The name of the file. Will be prefixed with 'pve-shm-'. Default is the VMID. Will be deleted when the VM is stopped.",
216 enum
=> [qw(ich9-intel-hda intel-hda AC97)],
217 description
=> "Configure an audio device."
221 enum
=> ['spice', 'none'],
224 description
=> "Driver backend for the audio device."
228 my $spice_enhancements_fmt = {
233 description
=> "Enable folder sharing via SPICE. Needs Spice-WebDAV daemon installed in the VM."
237 enum
=> ['off', 'all', 'filter'],
240 description
=> "Enable video streaming. Uses compression for detected video streams."
247 enum
=> ['/dev/urandom', '/dev/random', '/dev/hwrng'],
249 description
=> "The file on the host to gather entropy from. In most cases '/dev/urandom'"
250 ." should be preferred over '/dev/random' to avoid entropy-starvation issues on the"
251 ." host. Using urandom does *not* decrease security in any meaningful way, as it's"
252 ." still seeded from real entropy, and the bytes provided will most likely be mixed"
253 ." with real entropy on the guest as well. '/dev/hwrng' can be used to pass through"
254 ." a hardware RNG from the host.",
258 description
=> "Maximum bytes of entropy allowed to get injected into the guest every"
259 ." 'period' milliseconds. Prefer a lower value when using '/dev/random' as source. Use"
260 ." `0` to disable limiting (potentially dangerous!).",
263 # default is 1 KiB/s, provides enough entropy to the guest to avoid boot-starvation issues
264 # (e.g. systemd etc...) while allowing no chance of overwhelming the host, provided we're
265 # reading from /dev/urandom
270 description
=> "Every 'period' milliseconds the entropy-injection quota is reset, allowing"
271 ." the guest to retrieve another 'max_bytes' of entropy.",
277 my $meta_info_fmt = {
280 description
=> "The guest creation timestamp as UNIX epoch time",
286 description
=> "The QEMU (machine) version from the time this VM was created.",
287 pattern
=> '\d+(\.\d+)+',
296 description
=> "Specifies whether a VM will be started during system bootup.",
302 description
=> "Automatic restart after crash (currently ignored).",
307 type
=> 'string', format
=> 'pve-hotplug-features',
308 description
=> "Selectively enable hotplug features. This is a comma separated list of"
309 ." hotplug features: 'network', 'disk', 'cpu', 'memory', 'usb' and 'cloudinit'. Use '0' to disable"
310 ." hotplug completely. Using '1' as value is an alias for the default `network,disk,usb`."
311 ." USB hotplugging is possible for guests with machine version >= 7.1 and ostype l26 or"
313 default => 'network,disk,usb',
318 description
=> "Allow reboot. If set to '0' the VM exit on reboot.",
324 description
=> "Lock/unlock the VM.",
325 enum
=> [qw(backup clone create migrate rollback snapshot snapshot-delete suspending suspended)],
330 description
=> "Limit of CPU usage.",
331 verbose_description
=> "Limit of CPU usage.\n\nNOTE: If the computer has 2 CPUs, it has"
332 ." total of '2' CPU time. Value '0' indicates no CPU limit.",
340 description
=> "CPU weight for a VM, will be clamped to [1, 10000] in cgroup v2.",
341 verbose_description
=> "CPU weight for a VM. Argument is used in the kernel fair scheduler."
342 ." The larger the number is, the more CPU time this VM gets. Number is relative to"
343 ." weights of all the other running VMs.",
346 default => 'cgroup v1: 1024, cgroup v2: 100',
351 description
=> "Amount of RAM for the VM in MiB. This is the maximum available memory when"
352 ." you use the balloon device.",
359 description
=> "Amount of target RAM for the VM in MiB. Using zero disables the ballon driver.",
365 description
=> "Amount of memory shares for auto-ballooning. The larger the number is, the"
366 ." more memory this VM gets. Number is relative to weights of all other running VMs."
367 ." Using zero disables auto-ballooning. Auto-ballooning is done by pvestatd.",
375 description
=> "Keyboard layout for VNC server. This option is generally not required and"
376 ." is often better handled from within the guest OS.",
377 enum
=> PVE
::Tools
::kvmkeymaplist
(),
382 type
=> 'string', format
=> 'dns-name',
383 description
=> "Set a name for the VM. Only used on the configuration web interface.",
388 description
=> "SCSI controller model",
389 enum
=> [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
395 description
=> "Description for the VM. Shown in the web-interface VM's summary."
396 ." This is saved as comment inside the configuration file.",
397 maxLength
=> 1024 * 8,
402 enum
=> [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 win11 l24 l26 solaris)],
403 description
=> "Specify guest operating system.",
404 verbose_description
=> <<EODESC,
405 Specify guest operating system. This is used to enable special
406 optimization/features for specific operating systems:
409 other;; unspecified OS
410 wxp;; Microsoft Windows XP
411 w2k;; Microsoft Windows 2000
412 w2k3;; Microsoft Windows 2003
413 w2k8;; Microsoft Windows 2008
414 wvista;; Microsoft Windows Vista
415 win7;; Microsoft Windows 7
416 win8;; Microsoft Windows 8/2012/2012r2
417 win10;; Microsoft Windows 10/2016/2019
418 win11;; Microsoft Windows 11/2022
419 l24;; Linux 2.4 Kernel
420 l26;; Linux 2.6 - 6.X Kernel
421 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
426 type
=> 'string', format
=> 'pve-qm-boot',
427 description
=> "Specify guest boot order. Use the 'order=' sub-property as usage with no"
428 ." key or 'legacy=' is deprecated.",
432 type
=> 'string', format
=> 'pve-qm-bootdisk',
433 description
=> "Enable booting from specified disk. Deprecated: Use 'boot: order=foo;bar' instead.",
434 pattern
=> '(ide|sata|scsi|virtio)\d+',
439 description
=> "The number of CPUs. Please use option -sockets instead.",
446 description
=> "The number of CPU sockets.",
453 description
=> "The number of cores per socket.",
460 description
=> "Enable/disable NUMA.",
466 description
=> "Enable/disable hugepages memory.",
467 enum
=> [qw(any 2 1024)],
473 description
=> "Use together with hugepages. If enabled, hugepages will not not be deleted"
474 ." after VM shutdown and can be used for subsequent starts.",
479 description
=> "Number of hotplugged vcpus.",
486 description
=> "Enable/disable ACPI.",
491 description
=> "Enable/disable communication with the QEMU Guest Agent and its properties.",
493 format
=> $agent_fmt,
498 description
=> "Enable/disable KVM hardware virtualization.",
504 description
=> "Enable/disable time drift fix.",
510 description
=> "Set the real time clock (RTC) to local time. This is enabled by default if"
511 ." the `ostype` indicates a Microsoft Windows OS.",
516 description
=> "Freeze CPU at startup (use 'c' monitor command to start execution).",
520 type
=> 'string', format
=> $vga_fmt,
521 description
=> "Configure the VGA hardware.",
522 verbose_description
=> "Configure the VGA Hardware. If you want to use high resolution"
523 ." modes (>= 1280x1024x16) you may need to increase the vga memory option. Since QEMU"
524 ." 2.9 the default VGA display type is 'std' for all OS types besides some Windows"
525 ." versions (XP and older) which use 'cirrus'. The 'qxl' option enables the SPICE"
526 ." display server. For win* OS you can select how many independent displays you want,"
527 ." Linux guests can add displays them self.\nYou can also run without any graphic card,"
528 ." using a serial device as terminal.",
532 type
=> 'string', format
=> 'pve-qm-watchdog',
533 description
=> "Create a virtual hardware watchdog device.",
534 verbose_description
=> "Create a virtual hardware watchdog device. Once enabled (by a guest"
535 ." action), the watchdog must be periodically polled by an agent inside the guest or"
536 ." else the watchdog will reset the guest (or execute the respective action specified)",
541 typetext
=> "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
542 description
=> "Set the initial date of the real time clock. Valid format for date are:"
543 ."'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
544 pattern
=> '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
547 startup
=> get_standard_option
('pve-startup-order'),
551 description
=> "Enable/disable Template.",
557 description
=> "Arbitrary arguments passed to kvm.",
558 verbose_description
=> <<EODESCR,
559 Arbitrary arguments passed to kvm, for example:
561 args: -no-reboot -smbios 'type=0,vendor=FOO'
563 NOTE: this option is for experts only.
570 description
=> "Enable/disable the USB tablet device.",
571 verbose_description
=> "Enable/disable the USB tablet device. This device is usually needed"
572 ." to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with"
573 ." normal VNC clients. If you're running lots of console-only guests on one host, you"
574 ." may consider disabling this to save some context switches. This is turned off by"
575 ." default if you use spice (`qm set <vmid> --vga qxl`).",
580 description
=> "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
584 migrate_downtime
=> {
587 description
=> "Set maximum tolerated downtime (in seconds) for migrations.",
593 type
=> 'string', format
=> 'pve-qm-ide',
594 typetext
=> '<volume>',
595 description
=> "This is an alias for option -ide2",
599 description
=> "Emulated CPU type.",
601 format
=> 'pve-vm-cpu-conf',
603 parent
=> get_standard_option
('pve-snapshot-name', {
605 description
=> "Parent snapshot name. This is used internally, and should not be modified.",
609 description
=> "Timestamp for snapshots.",
615 type
=> 'string', format
=> 'pve-volume-id',
616 description
=> "Reference to a volume which stores the VM state. This is used internally"
619 vmstatestorage
=> get_standard_option
('pve-storage-id', {
620 description
=> "Default storage for VM state volumes/files.",
623 runningmachine
=> get_standard_option
('pve-qemu-machine', {
624 description
=> "Specifies the QEMU machine type of the running vm. This is used internally"
628 description
=> "Specifies the QEMU '-cpu' parameter of the running vm. This is used"
629 ." internally for snapshots.",
632 pattern
=> $PVE::QemuServer
::CPUConfig
::qemu_cmdline_cpu_re
,
633 format_description
=> 'QEMU -cpu parameter'
635 machine
=> get_standard_option
('pve-qemu-machine'),
637 description
=> "Virtual processor architecture. Defaults to the host.",
640 enum
=> [qw(x86_64 aarch64)],
643 description
=> "Specify SMBIOS type 1 fields.",
644 type
=> 'string', format
=> 'pve-qm-smbios1',
651 description
=> "Sets the protection flag of the VM. This will disable the remove VM and"
652 ." remove disk operations.",
658 enum
=> [ qw(seabios ovmf) ],
659 description
=> "Select BIOS implementation.",
660 default => 'seabios',
664 pattern
=> '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
665 format_description
=> 'UUID',
666 description
=> "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0'"
667 ." to disable explicitly.",
668 verbose_description
=> "The VM generation ID (vmgenid) device exposes a 128-bit integer"
669 ." value identifier to the guest OS. This allows to notify the guest operating system"
670 ." when the virtual machine is executed with a different configuration (e.g. snapshot"
671 ." execution or creation from a template). The guest operating system notices the"
672 ." change, and is then able to react as appropriate by marking its copies of"
673 ." distributed databases as dirty, re-initializing its random number generator, etc.\n"
674 ."Note that auto-creation only works when done through API/CLI create or update methods"
675 .", but not when manually editing the config file.",
676 default => "1 (autogenerated)",
681 format
=> 'pve-volume-id',
683 description
=> "Script that will be executed during various steps in the vms lifetime.",
687 format
=> $ivshmem_fmt,
688 description
=> "Inter-VM shared memory. Useful for direct communication between VMs, or to"
694 format
=> $audio_fmt,
695 description
=> "Configure a audio device, useful in combination with QXL/Spice.",
698 spice_enhancements
=> {
700 format
=> $spice_enhancements_fmt,
701 description
=> "Configure additional enhancements for SPICE.",
705 type
=> 'string', format
=> 'pve-tag-list',
706 description
=> 'Tags of the VM. This is only meta information.',
712 description
=> "Configure a VirtIO-based Random Number Generator.",
717 format
=> $meta_info_fmt,
718 description
=> "Some (read-only) meta-information about this guest.",
722 type
=> 'string', format
=> 'pve-cpuset',
723 description
=> "List of host cores used to execute guest processes, for example: 0,5,8-11",
732 description
=> 'Specify a custom file containing all meta data passed to the VM via"
733 ." cloud-init. This is provider specific meaning configdrive2 and nocloud differ.',
734 format
=> 'pve-volume-id',
735 format_description
=> 'volume',
740 description
=> 'To pass a custom file containing all network data to the VM via cloud-init.',
741 format
=> 'pve-volume-id',
742 format_description
=> 'volume',
747 description
=> 'To pass a custom file containing all user data to the VM via cloud-init.',
748 format
=> 'pve-volume-id',
749 format_description
=> 'volume',
754 description
=> 'To pass a custom file containing all vendor data to the VM via cloud-init.',
755 format
=> 'pve-volume-id',
756 format_description
=> 'volume',
759 PVE
::JSONSchema
::register_format
('pve-qm-cicustom', $cicustom_fmt);
761 my $confdesc_cloudinit = {
765 description
=> 'Specifies the cloud-init configuration format. The default depends on the'
766 .' configured operating system type (`ostype`. We use the `nocloud` format for Linux,'
767 .' and `configdrive2` for windows.',
768 enum
=> ['configdrive2', 'nocloud', 'opennebula'],
773 description
=> "cloud-init: User name to change ssh keys and password for instead of the"
774 ." image's configured default user.",
779 description
=> 'cloud-init: Password to assign the user. Using this is generally not'
780 .' recommended. Use ssh keys instead. Also note that older cloud-init versions do not'
781 .' support hashed passwords.',
786 description
=> 'cloud-init: do an automatic package upgrade after the first boot.'
791 description
=> 'cloud-init: Specify custom files to replace the automatically generated'
793 format
=> 'pve-qm-cicustom',
798 description
=> 'cloud-init: Sets DNS search domains for a container. Create will'
799 .' automatically use the setting from the host if neither searchdomain nor nameserver'
804 type
=> 'string', format
=> 'address-list',
805 description
=> 'cloud-init: Sets DNS server IP address for a container. Create will'
806 .' automatically use the setting from the host if neither searchdomain nor nameserver'
812 format
=> 'urlencoded',
813 description
=> "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
817 # what about other qemu settings ?
819 #machine => 'string',
832 ##soundhw => 'string',
834 while (my ($k, $v) = each %$confdesc) {
835 PVE
::JSONSchema
::register_standard_option
("pve-qm-$k", $v);
838 my $MAX_USB_DEVICES = 14;
840 my $MAX_SERIAL_PORTS = 4;
841 my $MAX_PARALLEL_PORTS = 3;
847 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
848 description
=> "CPUs accessing this NUMA node.",
849 format_description
=> "id[-id];...",
853 description
=> "Amount of memory this NUMA node provides.",
858 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
859 description
=> "Host NUMA nodes to use.",
860 format_description
=> "id[-id];...",
865 enum
=> [qw(preferred bind interleave)],
866 description
=> "NUMA allocation policy.",
870 PVE
::JSONSchema
::register_format
('pve-qm-numanode', $numa_fmt);
873 type
=> 'string', format
=> $numa_fmt,
874 description
=> "NUMA topology.",
876 PVE
::JSONSchema
::register_standard_option
("pve-qm-numanode", $numadesc);
878 for (my $i = 0; $i < $MAX_NUMA; $i++) {
879 $confdesc->{"numa$i"} = $numadesc;
882 my $nic_model_list = [
898 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
900 my $net_fmt_bridge_descr = <<__EOD__;
901 Bridge to attach the network device to. The Proxmox VE standard bridge
904 If you do not specify a bridge, we create a kvm user (NATed) network
905 device, which provides DHCP and DNS services. The following addresses
912 The DHCP server assign addresses to the guest starting from 10.0.2.15.
916 macaddr
=> get_standard_option
('mac-addr', {
917 description
=> "MAC address. That address must be unique withing your network. This is"
918 ." automatically generated if not specified.",
922 description
=> "Network Card Model. The 'virtio' model provides the best performance with"
923 ." very low CPU overhead. If your guest does not support this driver, it is usually"
924 ." best to use 'e1000'.",
925 enum
=> $nic_model_list,
928 (map { $_ => { keyAlias
=> 'model', alias
=> 'macaddr' }} @$nic_model_list),
929 bridge
=> get_standard_option
('pve-bridge-id', {
930 description
=> $net_fmt_bridge_descr,
935 minimum
=> 0, maximum
=> 64,
936 description
=> 'Number of packet queues to be used on the device.',
942 description
=> "Rate limit in mbps (megabytes per second) as floating point number.",
947 minimum
=> 1, maximum
=> 4094,
948 description
=> 'VLAN tag to apply to packets on this interface.',
953 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
954 description
=> 'VLAN trunks to pass through this interface.',
955 format_description
=> 'vlanid[;vlanid...]',
960 description
=> 'Whether this interface should be protected by the firewall.',
965 description
=> 'Whether this interface should be disconnected (like pulling the plug).',
970 minimum
=> 1, maximum
=> 65520,
971 description
=> "Force MTU, for VirtIO only. Set to '1' to use the bridge MTU",
978 type
=> 'string', format
=> $net_fmt,
979 description
=> "Specify network devices.",
982 PVE
::JSONSchema
::register_standard_option
("pve-qm-net", $netdesc);
987 format
=> 'pve-ipv4-config',
988 format_description
=> 'IPv4Format/CIDR',
989 description
=> 'IPv4 address in CIDR format.',
996 format_description
=> 'GatewayIPv4',
997 description
=> 'Default gateway for IPv4 traffic.',
1003 format
=> 'pve-ipv6-config',
1004 format_description
=> 'IPv6Format/CIDR',
1005 description
=> 'IPv6 address in CIDR format.',
1012 format_description
=> 'GatewayIPv6',
1013 description
=> 'Default gateway for IPv6 traffic.',
1018 PVE
::JSONSchema
::register_format
('pve-qm-ipconfig', $ipconfig_fmt);
1019 my $ipconfigdesc = {
1021 type
=> 'string', format
=> 'pve-qm-ipconfig',
1022 description
=> <<'EODESCR',
1023 cloud-init: Specify IP addresses and gateways for the corresponding interface.
1025 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
1027 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit
1028 gateway should be provided.
1029 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration. This requires
1030 cloud-init 19.4 or newer.
1032 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using
1036 PVE
::JSONSchema
::register_standard_option
("pve-qm-ipconfig", $netdesc);
1038 for (my $i = 0; $i < $MAX_NETS; $i++) {
1039 $confdesc->{"net$i"} = $netdesc;
1040 $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
1043 foreach my $key (keys %$confdesc_cloudinit) {
1044 $confdesc->{$key} = $confdesc_cloudinit->{$key};
1047 PVE
::JSONSchema
::register_format
('pve-cpuset', \
&pve_verify_cpuset
);
1048 sub pve_verify_cpuset
{
1049 my ($set_text, $noerr) = @_;
1051 my ($count, $members) = eval { PVE
::CpuSet
::parse_cpuset
($set_text) };
1055 die "unable to parse cpuset option\n";
1058 return PVE
::CpuSet-
>new($members)->short_string();
1061 PVE
::JSONSchema
::register_format
('pve-volume-id-or-qm-path', \
&verify_volume_id_or_qm_path
);
1062 sub verify_volume_id_or_qm_path
{
1063 my ($volid, $noerr) = @_;
1065 return $volid if $volid eq 'none' || $volid eq 'cdrom';
1067 return verify_volume_id_or_absolute_path
($volid, $noerr);
1070 PVE
::JSONSchema
::register_format
('pve-volume-id-or-absolute-path', \
&verify_volume_id_or_absolute_path
);
1071 sub verify_volume_id_or_absolute_path
{
1072 my ($volid, $noerr) = @_;
1074 return $volid if $volid =~ m
|^/|;
1076 $volid = eval { PVE
::JSONSchema
::check_format
('pve-volume-id', $volid, '') };
1087 type
=> 'string', format
=> 'pve-qm-usb-device',
1088 format_description
=> 'HOSTUSBDEVICE|spice',
1089 description
=> <<EODESCR,
1090 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1092 'bus-port(.port)*' (decimal numbers) or
1093 'vendor_id:product_id' (hexadeciaml numbers) or
1096 You can use the 'lsusb -t' command to list existing usb devices.
1098 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1099 machines - use with special care.
1101 The value 'spice' can be used to add a usb redirection devices for spice.
1107 description
=> "Specifies whether if given host option is a USB3 device or port."
1108 ." For modern guests (machine version >= 7.1 and ostype l26 and windows > 7), this flag"
1109 ." is irrelevant (all devices are plugged into a xhci controller).",
1116 type
=> 'string', format
=> $usb_fmt,
1117 description
=> "Configure an USB device (n is 0 to 4, for machine version >= 7.1 and ostype"
1118 ." l26 or windows > 7, n can be up to 14).",
1120 PVE
::JSONSchema
::register_standard_option
("pve-qm-usb", $usbdesc);
1125 pattern
=> '(/dev/.+|socket)',
1126 description
=> "Create a serial device inside the VM (n is 0 to 3)",
1127 verbose_description
=> <<EODESCR,
1128 Create a serial device inside the VM (n is 0 to 3), and pass through a
1129 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1130 host side (use 'qm terminal' to open a terminal connection).
1132 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines -
1133 use with special care.
1135 CAUTION: Experimental! User reported problems with this option.
1142 pattern
=> '/dev/parport\d+|/dev/usb/lp\d+',
1143 description
=> "Map host parallel devices (n is 0 to 2).",
1144 verbose_description
=> <<EODESCR,
1145 Map host parallel devices (n is 0 to 2).
1147 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1148 machines - use with special care.
1150 CAUTION: Experimental! User reported problems with this option.
1154 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
1155 $confdesc->{"parallel$i"} = $paralleldesc;
1158 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
1159 $confdesc->{"serial$i"} = $serialdesc;
1162 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
1163 $confdesc->{"hostpci$i"} = $PVE::QemuServer
::PCI
::hostpcidesc
;
1166 for my $key (keys %{$PVE::QemuServer
::Drive
::drivedesc_hash
}) {
1167 $confdesc->{$key} = $PVE::QemuServer
::Drive
::drivedesc_hash-
>{$key};
1170 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
1171 $confdesc->{"usb$i"} = $usbdesc;
1179 description
=> "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n)."
1180 . " Deprecated, use 'order=' instead.",
1181 pattern
=> '[acdn]{1,4}',
1182 format_description
=> "[acdn]{1,4}",
1184 # note: this is also the fallback if boot: is not given at all
1190 format
=> 'pve-qm-bootdev-list',
1191 format_description
=> "device[;device...]",
1192 description
=> <<EODESC,
1193 The guest will attempt to boot from devices in the order they appear here.
1195 Disks, optical drives and passed-through storage USB devices will be directly
1196 booted from, NICs will load PXE, and PCIe devices will either behave like disks
1197 (e.g. NVMe) or load an option ROM (e.g. RAID controller, hardware NIC).
1199 Note that only devices in this list will be marked as bootable and thus loaded
1200 by the guest firmware (BIOS/UEFI). If you require multiple disks for booting
1201 (e.g. software-raid), you need to specify all of them here.
1203 Overrides the deprecated 'legacy=[acdn]*' value when given.
1207 PVE
::JSONSchema
::register_format
('pve-qm-boot', $boot_fmt);
1209 PVE
::JSONSchema
::register_format
('pve-qm-bootdev', \
&verify_bootdev
);
1210 sub verify_bootdev
{
1211 my ($dev, $noerr) = @_;
1213 my $special = $dev =~ m/^efidisk/ || $dev =~ m/^tpmstate/;
1214 return $dev if PVE
::QemuServer
::Drive
::is_valid_drivename
($dev) && !$special;
1218 return 0 if $dev !~ m/^$base\d+$/;
1219 return 0 if !$confdesc->{$dev};
1223 return $dev if $check->("net");
1224 return $dev if $check->("usb");
1225 return $dev if $check->("hostpci");
1228 die "invalid boot device '$dev'\n";
1231 sub print_bootorder
{
1233 return "" if !@$devs;
1234 my $data = { order
=> join(';', @$devs) };
1235 return PVE
::JSONSchema
::print_property_string
($data, $boot_fmt);
1238 my $kvm_api_version = 0;
1241 return $kvm_api_version if $kvm_api_version;
1243 open my $fh, '<', '/dev/kvm' or return;
1245 # 0xae00 => KVM_GET_API_VERSION
1246 $kvm_api_version = ioctl($fh, 0xae00, 0);
1249 return $kvm_api_version;
1252 my $kvm_user_version = {};
1255 sub kvm_user_version
{
1258 $binary //= get_command_for_arch
(get_host_arch
()); # get the native arch by default
1259 my $st = stat($binary);
1261 my $cachedmtime = $kvm_mtime->{$binary} // -1;
1262 return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1263 $cachedmtime == $st->mtime;
1265 $kvm_user_version->{$binary} = 'unknown';
1266 $kvm_mtime->{$binary} = $st->mtime;
1270 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1271 $kvm_user_version->{$binary} = $2;
1275 eval { run_command
([$binary, '--version'], outfunc
=> $code); };
1278 return $kvm_user_version->{$binary};
1281 my sub extract_version
{
1282 my ($machine_type, $version) = @_;
1283 $version = kvm_user_version
() if !defined($version);
1284 return PVE
::QemuServer
::Machine
::extract_version
($machine_type, $version)
1287 sub kernel_has_vhost_net
{
1288 return -c
'/dev/vhost-net';
1293 return defined($confdesc->{$key});
1297 sub get_cdrom_path
{
1299 return $cdrom_path if defined($cdrom_path);
1301 $cdrom_path = first
{ -l
$_ } map { "/dev/cdrom$_" } ('', '1', '2');
1303 if (!defined($cdrom_path)) {
1304 log_warn
("no physical CD-ROM available, ignoring");
1312 my ($storecfg, $vmid, $cdrom) = @_;
1314 if ($cdrom eq 'cdrom') {
1315 return get_cdrom_path
();
1316 } elsif ($cdrom eq 'none') {
1318 } elsif ($cdrom =~ m
|^/|) {
1321 return PVE
::Storage
::path
($storecfg, $cdrom);
1325 # try to convert old style file names to volume IDs
1326 sub filename_to_volume_id
{
1327 my ($vmid, $file, $media) = @_;
1329 if (!($file eq 'none' || $file eq 'cdrom' ||
1330 $file =~ m
|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1332 return if $file =~ m
|/|;
1334 if ($media && $media eq 'cdrom') {
1335 $file = "local:iso/$file";
1337 $file = "local:$vmid/$file";
1344 sub verify_media_type
{
1345 my ($opt, $vtype, $media) = @_;
1350 if ($media eq 'disk') {
1352 } elsif ($media eq 'cdrom') {
1355 die "internal error";
1358 return if ($vtype eq $etype);
1360 raise_param_exc
({ $opt => "unexpected media type ($vtype != $etype)" });
1363 sub cleanup_drive_path
{
1364 my ($opt, $storecfg, $drive) = @_;
1366 # try to convert filesystem paths to volume IDs
1368 if (($drive->{file
} !~ m/^(cdrom|none)$/) &&
1369 ($drive->{file
} !~ m
|^/dev/.+|) &&
1370 ($drive->{file
} !~ m/^([^:]+):(.+)$/) &&
1371 ($drive->{file
} !~ m/^\d+$/)) {
1372 my ($vtype, $volid) = PVE
::Storage
::path_to_volume_id
($storecfg, $drive->{file
});
1373 raise_param_exc
({ $opt => "unable to associate path '$drive->{file}' to any storage"})
1375 $drive->{media
} = 'cdrom' if !$drive->{media
} && $vtype eq 'iso';
1376 verify_media_type
($opt, $vtype, $drive->{media
});
1377 $drive->{file
} = $volid;
1380 $drive->{media
} = 'cdrom' if !$drive->{media
} && $drive->{file
} =~ m/^(cdrom|none)$/;
1383 sub parse_hotplug_features
{
1388 return $res if $data eq '0';
1390 $data = $confdesc->{hotplug
}->{default} if $data eq '1';
1392 foreach my $feature (PVE
::Tools
::split_list
($data)) {
1393 if ($feature =~ m/^(network|disk|cpu|memory|usb|cloudinit)$/) {
1396 die "invalid hotplug feature '$feature'\n";
1402 PVE
::JSONSchema
::register_format
('pve-hotplug-features', \
&pve_verify_hotplug_features
);
1403 sub pve_verify_hotplug_features
{
1404 my ($value, $noerr) = @_;
1406 return $value if parse_hotplug_features
($value);
1410 die "unable to parse hotplug option\n";
1414 my($fh, $noerr) = @_;
1417 my $SG_GET_VERSION_NUM = 0x2282;
1419 my $versionbuf = "\x00" x
8;
1420 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1422 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1425 my $version = unpack("I", $versionbuf);
1426 if ($version < 30000) {
1427 die "scsi generic interface too old\n" if !$noerr;
1431 my $buf = "\x00" x
36;
1432 my $sensebuf = "\x00" x
8;
1433 my $cmd = pack("C x3 C x1", 0x12, 36);
1435 # see /usr/include/scsi/sg.h
1436 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";
1439 $sg_io_hdr_t, ord('S'), -3, length($cmd), length($sensebuf), 0, length($buf), $buf, $cmd, $sensebuf, 6000
1442 $ret = ioctl($fh, $SG_IO, $packet);
1444 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1448 my @res = unpack($sg_io_hdr_t, $packet);
1449 if ($res[17] || $res[18]) {
1450 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1455 $res->@{qw(type removable vendor product revision)} = unpack("C C x6 A8 A16 A4", $buf);
1457 $res->{removable
} = $res->{removable
} & 128 ?
1 : 0;
1458 $res->{type
} &= 0x1F;
1466 my $fh = IO
::File-
>new("+<$path") || return;
1467 my $res = scsi_inquiry
($fh, 1);
1473 sub print_tabletdevice_full
{
1474 my ($conf, $arch) = @_;
1476 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1478 # we use uhci for old VMs because tablet driver was buggy in older qemu
1480 if ($q35 || $arch eq 'aarch64') {
1486 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1489 sub print_keyboarddevice_full
{
1490 my ($conf, $arch) = @_;
1492 return if $arch ne 'aarch64';
1494 return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1497 my sub get_drive_id
{
1499 return "$drive->{interface}$drive->{index}";
1502 sub print_drivedevice_full
{
1503 my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1508 my $drive_id = get_drive_id
($drive);
1509 if ($drive->{interface
} eq 'virtio') {
1510 my $pciaddr = print_pci_addr
("$drive_id", $bridges, $arch, $machine_type);
1511 $device = "virtio-blk-pci,drive=drive-$drive_id,id=${drive_id}${pciaddr}";
1512 $device .= ",iothread=iothread-$drive_id" if $drive->{iothread
};
1513 } elsif ($drive->{interface
} eq 'scsi') {
1515 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
1516 my $unit = $drive->{index} % $maxdev;
1517 my $devicetype = 'hd';
1519 if (drive_is_cdrom
($drive)) {
1522 if ($drive->{file
} =~ m
|^/|) {
1523 $path = $drive->{file
};
1524 if (my $info = path_is_scsi
($path)) {
1525 if ($info->{type
} == 0 && $drive->{scsiblock
}) {
1526 $devicetype = 'block';
1527 } elsif ($info->{type
} == 1) { # tape
1528 $devicetype = 'generic';
1532 $path = PVE
::Storage
::path
($storecfg, $drive->{file
});
1535 # for compatibility only, we prefer scsi-hd (#2408, #2355, #2380)
1536 my $version = extract_version
($machine_type, kvm_user_version
());
1537 if ($path =~ m/^iscsi\:\/\
// &&
1538 !min_version
($version, 4, 1)) {
1539 $devicetype = 'generic';
1543 if (!$conf->{scsihw
} || $conf->{scsihw
} =~ m/^lsi/ || $conf->{scsihw
} eq 'pvscsi') {
1544 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit";
1546 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0"
1547 .",lun=$drive->{index}";
1549 $device .= ",drive=drive-$drive_id,id=$drive_id";
1551 if ($drive->{ssd
} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1552 $device .= ",rotation_rate=1";
1554 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1556 } elsif ($drive->{interface
} eq 'ide' || $drive->{interface
} eq 'sata') {
1557 my $maxdev = ($drive->{interface
} eq 'sata') ?
$PVE::QemuServer
::Drive
::MAX_SATA_DISKS
: 2;
1558 my $controller = int($drive->{index} / $maxdev);
1559 my $unit = $drive->{index} % $maxdev;
1560 my $devicetype = ($drive->{media
} && $drive->{media
} eq 'cdrom') ?
"cd" : "hd";
1562 $device = "ide-$devicetype";
1563 if ($drive->{interface
} eq 'ide') {
1564 $device .= ",bus=ide.$controller,unit=$unit";
1566 $device .= ",bus=ahci$controller.$unit";
1568 $device .= ",drive=drive-$drive_id,id=$drive_id";
1570 if ($devicetype eq 'hd') {
1571 if (my $model = $drive->{model
}) {
1572 $model = URI
::Escape
::uri_unescape
($model);
1573 $device .= ",model=$model";
1575 if ($drive->{ssd
}) {
1576 $device .= ",rotation_rate=1";
1579 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1580 } elsif ($drive->{interface
} eq 'usb') {
1582 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1584 die "unsupported interface type";
1587 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex
};
1589 if (my $serial = $drive->{serial
}) {
1590 $serial = URI
::Escape
::uri_unescape
($serial);
1591 $device .= ",serial=$serial";
1598 sub get_initiator_name
{
1601 my $fh = IO
::File-
>new('/etc/iscsi/initiatorname.iscsi') || return;
1602 while (defined(my $line = <$fh>)) {
1603 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1612 my sub storage_allows_io_uring_default
{
1613 my ($scfg, $cache_direct) = @_;
1615 # io_uring with cache mode writeback or writethrough on krbd will hang...
1616 return if $scfg && $scfg->{type
} eq 'rbd' && $scfg->{krbd
} && !$cache_direct;
1618 # io_uring with cache mode writeback or writethrough on LVM will hang, without cache only
1619 # sometimes, just plain disable...
1620 return if $scfg && $scfg->{type
} eq 'lvm';
1622 # io_uring causes problems when used with CIFS since kernel 5.15
1623 # Some discussion: https://www.spinics.net/lists/linux-cifs/msg26734.html
1624 return if $scfg && $scfg->{type
} eq 'cifs';
1629 my sub drive_uses_cache_direct
{
1630 my ($drive, $scfg) = @_;
1632 my $cache_direct = 0;
1634 if (my $cache = $drive->{cache
}) {
1635 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1636 } elsif (!drive_is_cdrom
($drive) && !($scfg && $scfg->{type
} eq 'btrfs' && !$scfg->{nocow
})) {
1640 return $cache_direct;
1643 sub print_drive_commandline_full
{
1644 my ($storecfg, $vmid, $drive, $pbs_name, $io_uring) = @_;
1647 my $volid = $drive->{file
};
1648 my $format = $drive->{format
};
1649 my $drive_id = get_drive_id
($drive);
1651 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1652 my $scfg = $storeid ? PVE
::Storage
::storage_config
($storecfg, $storeid) : undef;
1654 if (drive_is_cdrom
($drive)) {
1655 $path = get_iso_path
($storecfg, $vmid, $volid);
1656 die "$drive_id: cannot back cdrom drive with PBS snapshot\n" if $pbs_name;
1659 $path = PVE
::Storage
::path
($storecfg, $volid);
1660 $format //= qemu_img_format
($scfg, $volname);
1667 my $is_rbd = $path =~ m/^rbd:/;
1670 my @qemu_drive_options = qw(heads secs cyls trans media cache rerror werror aio discard);
1671 foreach my $o (@qemu_drive_options) {
1672 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1675 # snapshot only accepts on|off
1676 if (defined($drive->{snapshot
})) {
1677 my $v = $drive->{snapshot
} ?
'on' : 'off';
1678 $opts .= ",snapshot=$v";
1681 if (defined($drive->{ro
})) { # ro maps to QEMUs `readonly`, which accepts `on` or `off` only
1682 $opts .= ",readonly=" . ($drive->{ro
} ?
'on' : 'off');
1685 foreach my $type (['', '-total'], [_rd
=> '-read'], [_wr
=> '-write']) {
1686 my ($dir, $qmpname) = @$type;
1687 if (my $v = $drive->{"mbps$dir"}) {
1688 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1690 if (my $v = $drive->{"mbps${dir}_max"}) {
1691 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1693 if (my $v = $drive->{"bps${dir}_max_length"}) {
1694 $opts .= ",throttling.bps$qmpname-max-length=$v";
1696 if (my $v = $drive->{"iops${dir}"}) {
1697 $opts .= ",throttling.iops$qmpname=$v";
1699 if (my $v = $drive->{"iops${dir}_max"}) {
1700 $opts .= ",throttling.iops$qmpname-max=$v";
1702 if (my $v = $drive->{"iops${dir}_max_length"}) {
1703 $opts .= ",throttling.iops$qmpname-max-length=$v";
1708 $format = "rbd" if $is_rbd;
1709 die "$drive_id: Proxmox Backup Server backed drive cannot auto-detect the format\n"
1711 $opts .= ",format=alloc-track,file.driver=$format";
1713 $opts .= ",format=$format";
1716 my $cache_direct = drive_uses_cache_direct
($drive, $scfg);
1718 $opts .= ",cache=none" if !$drive->{cache
} && $cache_direct;
1720 if (!$drive->{aio
}) {
1721 if ($io_uring && storage_allows_io_uring_default
($scfg, $cache_direct)) {
1722 # io_uring supports all cache modes
1723 $opts .= ",aio=io_uring";
1725 # aio native works only with O_DIRECT
1727 $opts .= ",aio=native";
1729 $opts .= ",aio=threads";
1734 if (!drive_is_cdrom
($drive)) {
1736 if (defined($drive->{detect_zeroes
}) && !$drive->{detect_zeroes
}) {
1737 $detectzeroes = 'off';
1738 } elsif ($drive->{discard
}) {
1739 $detectzeroes = $drive->{discard
} eq 'on' ?
'unmap' : 'on';
1741 # This used to be our default with discard not being specified:
1742 $detectzeroes = 'on';
1745 # note: 'detect-zeroes' works per blockdev and we want it to persist
1746 # after the alloc-track is removed, so put it on 'file' directly
1747 my $dz_param = $pbs_name ?
"file.detect-zeroes" : "detect-zeroes";
1748 $opts .= ",$dz_param=$detectzeroes" if $detectzeroes;
1752 $opts .= ",backing=$pbs_name";
1753 $opts .= ",auto-remove=on";
1756 # my $file_param = $pbs_name ? "file.file.filename" : "file";
1757 my $file_param = "file";
1759 # non-rbd drivers require the underlying file to be a seperate block
1760 # node, so add a second .file indirection
1761 $file_param .= ".file" if !$is_rbd;
1762 $file_param .= ".filename";
1764 my $pathinfo = $path ?
"$file_param=$path," : '';
1766 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1769 sub print_pbs_blockdev
{
1770 my ($pbs_conf, $pbs_name) = @_;
1771 my $blockdev = "driver=pbs,node-name=$pbs_name,read-only=on";
1772 $blockdev .= ",repository=$pbs_conf->{repository}";
1773 $blockdev .= ",namespace=$pbs_conf->{namespace}" if $pbs_conf->{namespace
};
1774 $blockdev .= ",snapshot=$pbs_conf->{snapshot}";
1775 $blockdev .= ",archive=$pbs_conf->{archive}";
1776 $blockdev .= ",keyfile=$pbs_conf->{keyfile}" if $pbs_conf->{keyfile
};
1780 sub print_netdevice_full
{
1781 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version) = @_;
1783 my $device = $net->{model
};
1784 if ($net->{model
} eq 'virtio') {
1785 $device = 'virtio-net-pci';
1788 my $pciaddr = print_pci_addr
("$netid", $bridges, $arch, $machine_type);
1789 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1790 if ($net->{queues
} && $net->{queues
} > 1 && $net->{model
} eq 'virtio'){
1791 # Consider we have N queues, the number of vectors needed is 2 * N + 2, i.e., one per in
1792 # and out of each queue plus one config interrupt and control vector queue
1793 my $vectors = $net->{queues
} * 2 + 2;
1794 $tmpstr .= ",vectors=$vectors,mq=on";
1795 if (min_version
($machine_version, 7, 1)) {
1796 $tmpstr .= ",packed=on";
1800 if (min_version
($machine_version, 7, 1) && $net->{model
} eq 'virtio'){
1801 $tmpstr .= ",rx_queue_size=1024,tx_queue_size=1024";
1804 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex
} ;
1806 if (my $mtu = $net->{mtu
}) {
1807 if ($net->{model
} eq 'virtio' && $net->{bridge
}) {
1808 my $bridge_mtu = PVE
::Network
::read_bridge_mtu
($net->{bridge
});
1811 } elsif ($mtu < 576) {
1812 die "netdev $netid: MTU '$mtu' is smaller than the IP minimum MTU '576'\n";
1813 } elsif ($mtu > $bridge_mtu) {
1814 die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
1816 $tmpstr .= ",host_mtu=$mtu";
1818 warn "WARN: netdev $netid: ignoring MTU '$mtu', not using VirtIO or no bridge configured.\n";
1822 if ($use_old_bios_files) {
1824 if ($device eq 'virtio-net-pci') {
1825 $romfile = 'pxe-virtio.rom';
1826 } elsif ($device eq 'e1000') {
1827 $romfile = 'pxe-e1000.rom';
1828 } elsif ($device eq 'e1000e') {
1829 $romfile = 'pxe-e1000e.rom';
1830 } elsif ($device eq 'ne2k') {
1831 $romfile = 'pxe-ne2k_pci.rom';
1832 } elsif ($device eq 'pcnet') {
1833 $romfile = 'pxe-pcnet.rom';
1834 } elsif ($device eq 'rtl8139') {
1835 $romfile = 'pxe-rtl8139.rom';
1837 $tmpstr .= ",romfile=$romfile" if $romfile;
1843 sub print_netdev_full
{
1844 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
1847 if ($netid =~ m/^net(\d+)$/) {
1851 die "got strange net id '$i'\n" if $i >= ${MAX_NETS
};
1853 my $ifname = "tap${vmid}i$i";
1855 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1856 die "interface name '$ifname' is too long (max 15 character)\n"
1857 if length($ifname) >= 16;
1859 my $vhostparam = '';
1860 if (is_native
($arch)) {
1861 $vhostparam = ',vhost=on' if kernel_has_vhost_net
() && $net->{model
} eq 'virtio';
1864 my $vmname = $conf->{name
} || "vm$vmid";
1867 my $script = $hotplug ?
"pve-bridge-hotplug" : "pve-bridge";
1869 if ($net->{bridge
}) {
1870 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script"
1871 .",downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1873 $netdev = "type=user,id=$netid,hostname=$vmname";
1876 $netdev .= ",queues=$net->{queues}" if ($net->{queues
} && $net->{model
} eq 'virtio');
1882 'cirrus' => 'cirrus-vga',
1884 'vmware' => 'vmware-svga',
1885 'virtio' => 'virtio-vga',
1886 'virtio-gl' => 'virtio-vga-gl',
1889 sub print_vga_device
{
1890 my ($conf, $vga, $arch, $machine_version, $machine, $id, $qxlnum, $bridges) = @_;
1892 my $type = $vga_map->{$vga->{type
}};
1893 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
1894 $type = 'virtio-gpu';
1896 my $vgamem_mb = $vga->{memory
};
1898 my $max_outputs = '';
1900 $type = $id ?
'qxl' : 'qxl-vga';
1902 if (!$conf->{ostype
} || $conf->{ostype
} =~ m/^(?:l\d\d)|(?:other)$/) {
1903 # set max outputs so linux can have up to 4 qxl displays with one device
1904 if (min_version
($machine_version, 4, 1)) {
1905 $max_outputs = ",max_outputs=4";
1910 die "no devicetype for $vga->{type}\n" if !$type;
1914 if ($vga->{type
} =~ /^virtio/) {
1915 my $bytes = PVE
::Tools
::convert_size
($vgamem_mb, "mb" => "b");
1916 $memory = ",max_hostmem=$bytes";
1918 # from https://www.spice-space.org/multiple-monitors.html
1919 $memory = ",vgamem_mb=$vga->{memory}";
1920 my $ram = $vgamem_mb * 4;
1921 my $vram = $vgamem_mb * 2;
1922 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
1924 $memory = ",vgamem_mb=$vga->{memory}";
1926 } elsif ($qxlnum && $id) {
1927 $memory = ",ram_size=67108864,vram_size=33554432";
1931 if ($type eq 'VGA' && windows_version
($conf->{ostype
})) {
1932 $edidoff=",edid=off" if (!defined($conf->{bios
}) || $conf->{bios
} ne 'ovmf');
1935 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1936 my $vgaid = "vga" . ($id // '');
1938 if ($q35 && $vgaid eq 'vga') {
1939 # the first display uses pcie.0 bus on q35 machines
1940 $pciaddr = print_pcie_addr
($vgaid, $bridges, $arch, $machine);
1942 $pciaddr = print_pci_addr
($vgaid, $bridges, $arch, $machine);
1945 if ($vga->{type
} eq 'virtio-gl') {
1946 my $base = '/usr/lib/x86_64-linux-gnu/lib';
1947 die "missing libraries for '$vga->{type}' detected! Please install 'libgl1' and 'libegl1'\n"
1948 if !-e
"${base}EGL.so.1" || !-e
"${base}GL.so.1";
1950 die "no DRM render node detected (/dev/dri/renderD*), no GPU? - needed for '$vga->{type}' display\n"
1951 if !PVE
::Tools
::dir_glob_regex
('/dev/dri/', "renderD.*");
1954 return "$type,id=${vgaid}${memory}${max_outputs}${pciaddr}${edidoff}";
1957 sub parse_number_sets
{
1960 foreach my $part (split(/;/, $set)) {
1961 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1962 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1963 push @$res, [ $1, $2 ];
1965 die "invalid range: $part\n";
1974 my $res = parse_property_string
($numa_fmt, $data);
1975 $res->{cpus
} = parse_number_sets
($res->{cpus
}) if defined($res->{cpus
});
1976 $res->{hostnodes
} = parse_number_sets
($res->{hostnodes
}) if defined($res->{hostnodes
});
1980 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1982 my ($data, $disable_mac_autogen) = @_;
1984 my $res = eval { parse_property_string
($net_fmt, $data) };
1989 if (!defined($res->{macaddr
}) && !$disable_mac_autogen) {
1990 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
1991 $res->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
});
1996 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
1997 sub parse_ipconfig
{
2000 my $res = eval { parse_property_string
($ipconfig_fmt, $data) };
2006 if ($res->{gw
} && !$res->{ip
}) {
2007 warn 'gateway specified without specifying an IP address';
2010 if ($res->{gw6
} && !$res->{ip6
}) {
2011 warn 'IPv6 gateway specified without specifying an IPv6 address';
2014 if ($res->{gw
} && $res->{ip
} eq 'dhcp') {
2015 warn 'gateway specified together with DHCP';
2018 if ($res->{gw6
} && $res->{ip6
} !~ /^$IPV6RE/) {
2020 warn "IPv6 gateway specified together with $res->{ip6} address";
2024 if (!$res->{ip
} && !$res->{ip6
}) {
2025 return { ip
=> 'dhcp', ip6
=> 'dhcp' };
2034 return PVE
::JSONSchema
::print_property_string
($net, $net_fmt);
2037 sub add_random_macs
{
2038 my ($settings) = @_;
2040 foreach my $opt (keys %$settings) {
2041 next if $opt !~ m/^net(\d+)$/;
2042 my $net = parse_net
($settings->{$opt});
2044 $settings->{$opt} = print_net
($net);
2048 sub vm_is_volid_owner
{
2049 my ($storecfg, $vmid, $volid) = @_;
2051 if ($volid !~ m
|^/|) {
2053 eval { ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid); };
2054 if ($owner && ($owner == $vmid)) {
2062 sub vmconfig_register_unused_drive
{
2063 my ($storecfg, $vmid, $conf, $drive) = @_;
2065 if (drive_is_cloudinit
($drive)) {
2066 eval { PVE
::Storage
::vdisk_free
($storecfg, $drive->{file
}) };
2068 delete $conf->{cloudinit
};
2069 } elsif (!drive_is_cdrom
($drive)) {
2070 my $volid = $drive->{file
};
2071 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
2072 PVE
::QemuConfig-
>add_unused_volume($conf, $volid, $vmid);
2077 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2081 pattern
=> '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2082 format_description
=> 'UUID',
2083 description
=> "Set SMBIOS1 UUID.",
2088 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2089 format_description
=> 'Base64 encoded string',
2090 description
=> "Set SMBIOS1 version.",
2095 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2096 format_description
=> 'Base64 encoded string',
2097 description
=> "Set SMBIOS1 serial number.",
2102 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2103 format_description
=> 'Base64 encoded string',
2104 description
=> "Set SMBIOS1 manufacturer.",
2109 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2110 format_description
=> 'Base64 encoded string',
2111 description
=> "Set SMBIOS1 product ID.",
2116 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2117 format_description
=> 'Base64 encoded string',
2118 description
=> "Set SMBIOS1 SKU string.",
2123 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2124 format_description
=> 'Base64 encoded string',
2125 description
=> "Set SMBIOS1 family string.",
2130 description
=> 'Flag to indicate that the SMBIOS values are base64 encoded',
2138 my $res = eval { parse_property_string
($smbios1_fmt, $data) };
2145 return PVE
::JSONSchema
::print_property_string
($smbios1, $smbios1_fmt);
2148 PVE
::JSONSchema
::register_format
('pve-qm-smbios1', $smbios1_fmt);
2150 sub parse_watchdog
{
2155 my $res = eval { parse_property_string
($watchdog_fmt, $value) };
2160 sub parse_guest_agent
{
2163 return {} if !defined($conf->{agent
});
2165 my $res = eval { parse_property_string
($agent_fmt, $conf->{agent
}) };
2168 # if the agent is disabled ignore the other potentially set properties
2169 return {} if !$res->{enabled
};
2174 my ($conf, $key) = @_;
2175 return undef if !defined($conf->{agent
});
2177 my $agent = parse_guest_agent
($conf);
2178 return $agent->{$key};
2184 return {} if !$value;
2185 my $res = eval { parse_property_string
($vga_fmt, $value) };
2195 my $res = eval { parse_property_string
($rng_fmt, $value) };
2200 sub parse_meta_info
{
2205 my $res = eval { parse_property_string
($meta_info_fmt, $value) };
2210 sub new_meta_info_string
{
2211 my () = @_; # for now do not allow to override any value
2213 return PVE
::JSONSchema
::print_property_string
(
2215 'creation-qemu' => kvm_user_version
(),
2216 ctime
=> "". int(time()),
2222 sub qemu_created_version_fixups
{
2223 my ($conf, $forcemachine, $kvmver) = @_;
2225 my $meta = parse_meta_info
($conf->{meta
}) // {};
2226 my $forced_vers = PVE
::QemuServer
::Machine
::extract_version
($forcemachine);
2228 # check if we need to apply some handling for VMs that always use the latest machine version but
2229 # had a machine version transition happen that affected HW such that, e.g., an OS config change
2230 # would be required (we do not want to pin machine version for non-windows OS type)
2232 (!defined($conf->{machine
}) || $conf->{machine
} =~ m/^(?:pc|q35|virt)$/) # non-versioned machine
2233 && (!defined($meta->{'creation-qemu'}) || !min_version
($meta->{'creation-qemu'}, 6, 1)) # created before 6.1
2234 && (!$forced_vers || min_version
($forced_vers, 6, 1)) # handle snapshot-rollback/migrations
2235 && min_version
($kvmver, 6, 1) # only need to apply the change since 6.1
2237 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
2238 if ($q35 && $conf->{ostype
} && $conf->{ostype
} eq 'l26') {
2239 # this changed to default-on in Q 6.1 for q35 machines, it will mess with PCI slot view
2240 # and thus with the predictable interface naming of systemd
2241 return ['-global', 'ICH9-LPC.acpi-pci-hotplug-with-bridge-support=off'];
2247 PVE
::JSONSchema
::register_format
('pve-qm-usb-device', \
&verify_usb_device
);
2248 sub verify_usb_device
{
2249 my ($value, $noerr) = @_;
2251 return $value if parse_usb_device
($value);
2255 die "unable to parse usb device\n";
2258 # add JSON properties for create and set function
2259 sub json_config_properties
{
2260 my ($prop, $with_disk_alloc) = @_;
2262 my $skip_json_config_opts = {
2266 runningmachine
=> 1,
2271 foreach my $opt (keys %$confdesc) {
2272 next if $skip_json_config_opts->{$opt};
2274 if ($with_disk_alloc && is_valid_drivename
($opt)) {
2275 $prop->{$opt} = $PVE::QemuServer
::Drive
::drivedesc_hash_with_alloc-
>{$opt};
2277 $prop->{$opt} = $confdesc->{$opt};
2284 # Properties that we can read from an OVF file
2285 sub json_ovf_properties
{
2288 for my $device (PVE
::QemuServer
::Drive
::valid_drive_names
()) {
2289 $prop->{$device} = {
2291 format
=> 'pve-volume-id-or-absolute-path',
2292 description
=> "Disk image that gets imported to $device",
2299 description
=> "The number of CPU cores.",
2304 description
=> "Amount of RAM for the VM in MB.",
2309 description
=> "Name of the VM.",
2316 # return copy of $confdesc_cloudinit to generate documentation
2317 sub cloudinit_config_properties
{
2319 return dclone
($confdesc_cloudinit);
2322 sub cloudinit_pending_properties
{
2324 map { $_ => 1 } keys $confdesc_cloudinit->%*,
2327 $p->{"net$_"} = 1 for 0..($MAX_NETS-1);
2332 my ($key, $value) = @_;
2334 die "unknown setting '$key'\n" if !$confdesc->{$key};
2336 my $type = $confdesc->{$key}->{type
};
2338 if (!defined($value)) {
2339 die "got undefined value\n";
2342 if ($value =~ m/[\n\r]/) {
2343 die "property contains a line feed\n";
2346 if ($type eq 'boolean') {
2347 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2348 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2349 die "type check ('boolean') failed - got '$value'\n";
2350 } elsif ($type eq 'integer') {
2351 return int($1) if $value =~ m/^(\d+)$/;
2352 die "type check ('integer') failed - got '$value'\n";
2353 } elsif ($type eq 'number') {
2354 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2355 die "type check ('number') failed - got '$value'\n";
2356 } elsif ($type eq 'string') {
2357 if (my $fmt = $confdesc->{$key}->{format
}) {
2358 PVE
::JSONSchema
::check_format
($fmt, $value);
2361 $value =~ s/^\"(.*)\"$/$1/;
2364 die "internal error"
2369 my ($storecfg, $vmid, $skiplock, $replacement_conf, $purge_unreferenced) = @_;
2371 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2373 if (!$skiplock && !PVE
::QemuConfig-
>has_lock($conf, 'suspended')) {
2374 PVE
::QemuConfig-
>check_lock($conf);
2377 if ($conf->{template
}) {
2378 # check if any base image is still used by a linked clone
2379 PVE
::QemuConfig-
>foreach_volume_full($conf, { include_unused
=> 1 }, sub {
2380 my ($ds, $drive) = @_;
2381 return if drive_is_cdrom
($drive);
2383 my $volid = $drive->{file
};
2384 return if !$volid || $volid =~ m
|^/|;
2386 die "base volume '$volid' is still in use by linked cloned\n"
2387 if PVE
::Storage
::volume_is_base_and_used
($storecfg, $volid);
2393 my $remove_owned_drive = sub {
2394 my ($ds, $drive) = @_;
2395 return if drive_is_cdrom
($drive, 1);
2397 my $volid = $drive->{file
};
2398 return if !$volid || $volid =~ m
|^/|;
2399 return if $volids->{$volid};
2401 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
2402 return if !$path || !$owner || ($owner != $vmid);
2404 $volids->{$volid} = 1;
2405 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2406 warn "Could not remove disk '$volid', check manually: $@" if $@;
2409 # only remove disks owned by this VM (referenced in the config)
2410 my $include_opts = {
2411 include_unused
=> 1,
2412 extra_keys
=> ['vmstate'],
2414 PVE
::QemuConfig-
>foreach_volume_full($conf, $include_opts, $remove_owned_drive);
2416 for my $snap (values %{$conf->{snapshots
}}) {
2417 next if !defined($snap->{vmstate
});
2418 my $drive = PVE
::QemuConfig-
>parse_volume('vmstate', $snap->{vmstate
}, 1);
2419 next if !defined($drive);
2420 $remove_owned_drive->('vmstate', $drive);
2423 PVE
::QemuConfig-
>foreach_volume_full($conf->{pending
}, $include_opts, $remove_owned_drive);
2425 if ($purge_unreferenced) { # also remove unreferenced disk
2426 my $vmdisks = PVE
::Storage
::vdisk_list
($storecfg, undef, $vmid, undef, 'images');
2427 PVE
::Storage
::foreach_volid
($vmdisks, sub {
2428 my ($volid, $sid, $volname, $d) = @_;
2429 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2434 if (defined $replacement_conf) {
2435 PVE
::QemuConfig-
>write_config($vmid, $replacement_conf);
2437 PVE
::QemuConfig-
>destroy_config($vmid);
2441 sub parse_vm_config
{
2442 my ($filename, $raw, $strict) = @_;
2444 return if !defined($raw);
2447 digest
=> Digest
::SHA
::sha1_hex
($raw),
2453 my $handle_error = sub {
2463 $filename =~ m
|/qemu-server/(\d
+)\
.conf
$|
2464 || die "got strange filename '$filename'";
2470 my $finish_description = sub {
2471 if (defined($descr)) {
2473 $conf->{description
} = $descr;
2479 my @lines = split(/\n/, $raw);
2480 foreach my $line (@lines) {
2481 next if $line =~ m/^\s*$/;
2483 if ($line =~ m/^\[PENDING\]\s*$/i) {
2484 $section = 'pending';
2485 $finish_description->();
2486 $conf = $res->{$section} = {};
2488 } elsif ($line =~ m/^\[special:cloudinit\]\s*$/i) {
2489 $section = 'cloudinit';
2490 $finish_description->();
2491 $conf = $res->{$section} = {};
2494 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2496 $finish_description->();
2497 $conf = $res->{snapshots
}->{$section} = {};
2501 if ($line =~ m/^\#(.*)$/) {
2502 $descr = '' if !defined($descr);
2503 $descr .= PVE
::Tools
::decode_text
($1) . "\n";
2507 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2508 $descr = '' if !defined($descr);
2509 $descr .= PVE
::Tools
::decode_text
($2);
2510 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2511 $conf->{snapstate
} = $1;
2512 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2515 $conf->{$key} = $value;
2516 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2518 if ($section eq 'pending') {
2519 $conf->{delete} = $value; # we parse this later
2521 $handle_error->("vm $vmid - property 'delete' is only allowed in [PENDING]\n");
2523 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2526 if ($section eq 'cloudinit') {
2527 # ignore validation only used for informative purpose
2528 $conf->{$key} = $value;
2531 eval { $value = check_type
($key, $value); };
2533 $handle_error->("vm $vmid - unable to parse value of '$key' - $@");
2535 $key = 'ide2' if $key eq 'cdrom';
2536 my $fmt = $confdesc->{$key}->{format
};
2537 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2538 my $v = parse_drive
($key, $value);
2539 if (my $volid = filename_to_volume_id
($vmid, $v->{file
}, $v->{media
})) {
2540 $v->{file
} = $volid;
2541 $value = print_drive
($v);
2543 $handle_error->("vm $vmid - unable to parse value of '$key'\n");
2548 $conf->{$key} = $value;
2551 $handle_error->("vm $vmid - unable to parse config: $line\n");
2555 $finish_description->();
2556 delete $res->{snapstate
}; # just to be sure
2561 sub write_vm_config
{
2562 my ($filename, $conf) = @_;
2564 delete $conf->{snapstate
}; # just to be sure
2566 if ($conf->{cdrom
}) {
2567 die "option ide2 conflicts with cdrom\n" if $conf->{ide2
};
2568 $conf->{ide2
} = $conf->{cdrom
};
2569 delete $conf->{cdrom
};
2572 # we do not use 'smp' any longer
2573 if ($conf->{sockets
}) {
2574 delete $conf->{smp
};
2575 } elsif ($conf->{smp
}) {
2576 $conf->{sockets
} = $conf->{smp
};
2577 delete $conf->{cores
};
2578 delete $conf->{smp
};
2581 my $used_volids = {};
2583 my $cleanup_config = sub {
2584 my ($cref, $pending, $snapname) = @_;
2586 foreach my $key (keys %$cref) {
2587 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2588 $key eq 'snapstate' || $key eq 'pending' || $key eq 'cloudinit';
2589 my $value = $cref->{$key};
2590 if ($key eq 'delete') {
2591 die "propertry 'delete' is only allowed in [PENDING]\n"
2593 # fixme: check syntax?
2596 eval { $value = check_type
($key, $value); };
2597 die "unable to parse value of '$key' - $@" if $@;
2599 $cref->{$key} = $value;
2601 if (!$snapname && is_valid_drivename
($key)) {
2602 my $drive = parse_drive
($key, $value);
2603 $used_volids->{$drive->{file
}} = 1 if $drive && $drive->{file
};
2608 &$cleanup_config($conf);
2610 &$cleanup_config($conf->{pending
}, 1);
2612 foreach my $snapname (keys %{$conf->{snapshots
}}) {
2613 die "internal error: snapshot name '$snapname' is forbidden" if lc($snapname) eq 'pending';
2614 &$cleanup_config($conf->{snapshots
}->{$snapname}, undef, $snapname);
2617 # remove 'unusedX' settings if we re-add a volume
2618 foreach my $key (keys %$conf) {
2619 my $value = $conf->{$key};
2620 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2621 delete $conf->{$key};
2625 my $generate_raw_config = sub {
2626 my ($conf, $pending) = @_;
2630 # add description as comment to top of file
2631 if (defined(my $descr = $conf->{description
})) {
2633 foreach my $cl (split(/\n/, $descr)) {
2634 $raw .= '#' . PVE
::Tools
::encode_text
($cl) . "\n";
2637 $raw .= "#\n" if $pending;
2641 foreach my $key (sort keys %$conf) {
2642 next if $key =~ /^(digest|description|pending|cloudinit|snapshots)$/;
2643 $raw .= "$key: $conf->{$key}\n";
2648 my $raw = &$generate_raw_config($conf);
2650 if (scalar(keys %{$conf->{pending
}})){
2651 $raw .= "\n[PENDING]\n";
2652 $raw .= &$generate_raw_config($conf->{pending
}, 1);
2655 if (scalar(keys %{$conf->{cloudinit
}}) && PVE
::QemuConfig-
>has_cloudinit($conf)){
2656 $raw .= "\n[special:cloudinit]\n";
2657 $raw .= &$generate_raw_config($conf->{cloudinit
});
2660 foreach my $snapname (sort keys %{$conf->{snapshots
}}) {
2661 $raw .= "\n[$snapname]\n";
2662 $raw .= &$generate_raw_config($conf->{snapshots
}->{$snapname});
2672 # we use static defaults from our JSON schema configuration
2673 foreach my $key (keys %$confdesc) {
2674 if (defined(my $default = $confdesc->{$key}->{default})) {
2675 $res->{$key} = $default;
2683 my $vmlist = PVE
::Cluster
::get_vmlist
();
2685 return $res if !$vmlist || !$vmlist->{ids
};
2686 my $ids = $vmlist->{ids
};
2687 my $nodename = nodename
();
2689 foreach my $vmid (keys %$ids) {
2690 my $d = $ids->{$vmid};
2691 next if !$d->{node
} || $d->{node
} ne $nodename;
2692 next if !$d->{type
} || $d->{type
} ne 'qemu';
2693 $res->{$vmid}->{exists} = 1;
2698 # test if VM uses local resources (to prevent migration)
2699 sub check_local_resources
{
2700 my ($conf, $noerr) = @_;
2704 push @loc_res, "hostusb" if $conf->{hostusb
}; # old syntax
2705 push @loc_res, "hostpci" if $conf->{hostpci
}; # old syntax
2707 push @loc_res, "ivshmem" if $conf->{ivshmem
};
2709 foreach my $k (keys %$conf) {
2710 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2711 # sockets are safe: they will recreated be on the target side post-migrate
2712 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2713 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2716 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2721 # check if used storages are available on all nodes (use by migrate)
2722 sub check_storage_availability
{
2723 my ($storecfg, $conf, $node) = @_;
2725 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2726 my ($ds, $drive) = @_;
2728 my $volid = $drive->{file
};
2731 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2734 # check if storage is available on both nodes
2735 my $scfg = PVE
::Storage
::storage_check_enabled
($storecfg, $sid);
2736 PVE
::Storage
::storage_check_enabled
($storecfg, $sid, $node);
2738 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $volid);
2740 die "$volid: content type '$vtype' is not available on storage '$sid'\n"
2741 if !$scfg->{content
}->{$vtype};
2745 # list nodes where all VM images are available (used by has_feature API)
2747 my ($conf, $storecfg) = @_;
2749 my $nodelist = PVE
::Cluster
::get_nodelist
();
2750 my $nodehash = { map { $_ => 1 } @$nodelist };
2751 my $nodename = nodename
();
2753 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2754 my ($ds, $drive) = @_;
2756 my $volid = $drive->{file
};
2759 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2761 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2762 if ($scfg->{disable
}) {
2764 } elsif (my $avail = $scfg->{nodes
}) {
2765 foreach my $node (keys %$nodehash) {
2766 delete $nodehash->{$node} if !$avail->{$node};
2768 } elsif (!$scfg->{shared
}) {
2769 foreach my $node (keys %$nodehash) {
2770 delete $nodehash->{$node} if $node ne $nodename
2779 sub check_local_storage_availability
{
2780 my ($conf, $storecfg) = @_;
2782 my $nodelist = PVE
::Cluster
::get_nodelist
();
2783 my $nodehash = { map { $_ => {} } @$nodelist };
2785 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2786 my ($ds, $drive) = @_;
2788 my $volid = $drive->{file
};
2791 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2793 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2795 if ($scfg->{disable
}) {
2796 foreach my $node (keys %$nodehash) {
2797 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2799 } elsif (my $avail = $scfg->{nodes
}) {
2800 foreach my $node (keys %$nodehash) {
2801 if (!$avail->{$node}) {
2802 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2809 foreach my $node (values %$nodehash) {
2810 if (my $unavail = $node->{unavailable_storages
}) {
2811 $node->{unavailable_storages
} = [ sort keys %$unavail ];
2818 # Compat only, use assert_config_exists_on_node and vm_running_locally where possible
2820 my ($vmid, $nocheck, $node) = @_;
2822 # $nocheck is set when called during a migration, in which case the config
2823 # file might still or already reside on the *other* node
2824 # - because rename has already happened, and current node is source
2825 # - because rename hasn't happened yet, and current node is target
2826 # - because rename has happened, current node is target, but hasn't yet
2828 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid, $node) if !$nocheck;
2829 return PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
2834 my $vzlist = config_list
();
2836 my $fd = IO
::Dir-
>new($PVE::QemuServer
::Helpers
::var_run_tmpdir
) || return $vzlist;
2838 while (defined(my $de = $fd->read)) {
2839 next if $de !~ m/^(\d+)\.pid$/;
2841 next if !defined($vzlist->{$vmid});
2842 if (my $pid = check_running
($vmid)) {
2843 $vzlist->{$vmid}->{pid
} = $pid;
2850 our $vmstatus_return_properties = {
2851 vmid
=> get_standard_option
('pve-vmid'),
2853 description
=> "QEMU process status.",
2855 enum
=> ['stopped', 'running'],
2858 description
=> "Maximum memory in bytes.",
2861 renderer
=> 'bytes',
2864 description
=> "Root disk size in bytes.",
2867 renderer
=> 'bytes',
2870 description
=> "VM name.",
2875 description
=> "QEMU QMP agent status.",
2880 description
=> "PID of running qemu process.",
2885 description
=> "Uptime.",
2888 renderer
=> 'duration',
2891 description
=> "Maximum usable CPUs.",
2896 description
=> "The current config lock, if any.",
2901 description
=> "The current configured tags, if any",
2905 'running-machine' => {
2906 description
=> "The currently running machine type (if running).",
2911 description
=> "The currently running QEMU version (if running).",
2917 my $last_proc_pid_stat;
2919 # get VM status information
2920 # This must be fast and should not block ($full == false)
2921 # We only query KVM using QMP if $full == true (this can be slow)
2923 my ($opt_vmid, $full) = @_;
2927 my $storecfg = PVE
::Storage
::config
();
2929 my $list = vzlist
();
2930 my $defaults = load_defaults
();
2932 my ($uptime) = PVE
::ProcFSTools
::read_proc_uptime
(1);
2934 my $cpucount = $cpuinfo->{cpus
} || 1;
2936 foreach my $vmid (keys %$list) {
2937 next if $opt_vmid && ($vmid ne $opt_vmid);
2939 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2941 my $d = { vmid
=> int($vmid) };
2942 $d->{pid
} = int($list->{$vmid}->{pid
}) if $list->{$vmid}->{pid
};
2944 # fixme: better status?
2945 $d->{status
} = $list->{$vmid}->{pid
} ?
'running' : 'stopped';
2947 my $size = PVE
::QemuServer
::Drive
::bootdisk_size
($storecfg, $conf);
2948 if (defined($size)) {
2949 $d->{disk
} = 0; # no info available
2950 $d->{maxdisk
} = $size;
2956 $d->{cpus
} = ($conf->{sockets
} || $defaults->{sockets
})
2957 * ($conf->{cores
} || $defaults->{cores
});
2958 $d->{cpus
} = $cpucount if $d->{cpus
} > $cpucount;
2959 $d->{cpus
} = $conf->{vcpus
} if $conf->{vcpus
};
2961 $d->{name
} = $conf->{name
} || "VM $vmid";
2962 $d->{maxmem
} = $conf->{memory
} ?
$conf->{memory
}*(1024*1024)
2963 : $defaults->{memory
}*(1024*1024);
2965 if ($conf->{balloon
}) {
2966 $d->{balloon_min
} = $conf->{balloon
}*(1024*1024);
2967 $d->{shares
} = defined($conf->{shares
}) ?
$conf->{shares
}
2968 : $defaults->{shares
};
2979 $d->{diskwrite
} = 0;
2981 $d->{template
} = 1 if PVE
::QemuConfig-
>is_template($conf);
2983 $d->{serial
} = 1 if conf_has_serial
($conf);
2984 $d->{lock} = $conf->{lock} if $conf->{lock};
2985 $d->{tags
} = $conf->{tags
} if defined($conf->{tags
});
2990 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
2991 foreach my $dev (keys %$netdev) {
2992 next if $dev !~ m/^tap([1-9]\d*)i/;
2994 my $d = $res->{$vmid};
2997 $d->{netout
} += $netdev->{$dev}->{receive
};
2998 $d->{netin
} += $netdev->{$dev}->{transmit
};
3001 $d->{nics
}->{$dev}->{netout
} = int($netdev->{$dev}->{receive
});
3002 $d->{nics
}->{$dev}->{netin
} = int($netdev->{$dev}->{transmit
});
3007 my $ctime = gettimeofday
;
3009 foreach my $vmid (keys %$list) {
3011 my $d = $res->{$vmid};
3012 my $pid = $d->{pid
};
3015 my $pstat = PVE
::ProcFSTools
::read_proc_pid_stat
($pid);
3016 next if !$pstat; # not running
3018 my $used = $pstat->{utime} + $pstat->{stime
};
3020 $d->{uptime
} = int(($uptime - $pstat->{starttime
})/$cpuinfo->{user_hz
});
3022 if ($pstat->{vsize
}) {
3023 $d->{mem
} = int(($pstat->{rss
}/$pstat->{vsize
})*$d->{maxmem
});
3026 my $old = $last_proc_pid_stat->{$pid};
3028 $last_proc_pid_stat->{$pid} = {
3036 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
3038 if ($dtime > 1000) {
3039 my $dutime = $used - $old->{used
};
3041 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
3042 $last_proc_pid_stat->{$pid} = {
3048 $d->{cpu
} = $old->{cpu
};
3052 return $res if !$full;
3054 my $qmpclient = PVE
::QMPClient-
>new();
3056 my $ballooncb = sub {
3057 my ($vmid, $resp) = @_;
3059 my $info = $resp->{'return'};
3060 return if !$info->{max_mem
};
3062 my $d = $res->{$vmid};
3064 # use memory assigned to VM
3065 $d->{maxmem
} = $info->{max_mem
};
3066 $d->{balloon
} = $info->{actual
};
3068 if (defined($info->{total_mem
}) && defined($info->{free_mem
})) {
3069 $d->{mem
} = $info->{total_mem
} - $info->{free_mem
};
3070 $d->{freemem
} = $info->{free_mem
};
3073 $d->{ballooninfo
} = $info;
3076 my $blockstatscb = sub {
3077 my ($vmid, $resp) = @_;
3078 my $data = $resp->{'return'} || [];
3079 my $totalrdbytes = 0;
3080 my $totalwrbytes = 0;
3082 for my $blockstat (@$data) {
3083 $totalrdbytes = $totalrdbytes + $blockstat->{stats
}->{rd_bytes
};
3084 $totalwrbytes = $totalwrbytes + $blockstat->{stats
}->{wr_bytes
};
3086 $blockstat->{device
} =~ s/drive-//;
3087 $res->{$vmid}->{blockstat
}->{$blockstat->{device
}} = $blockstat->{stats
};
3089 $res->{$vmid}->{diskread
} = $totalrdbytes;
3090 $res->{$vmid}->{diskwrite
} = $totalwrbytes;
3093 my $machinecb = sub {
3094 my ($vmid, $resp) = @_;
3095 my $data = $resp->{'return'} || [];
3097 $res->{$vmid}->{'running-machine'} =
3098 PVE
::QemuServer
::Machine
::current_from_query_machines
($data);
3101 my $versioncb = sub {
3102 my ($vmid, $resp) = @_;
3103 my $data = $resp->{'return'} // {};
3104 my $version = 'unknown';
3106 if (my $v = $data->{qemu
}) {
3107 $version = $v->{major
} . "." . $v->{minor
} . "." . $v->{micro
};
3110 $res->{$vmid}->{'running-qemu'} = $version;
3113 my $statuscb = sub {
3114 my ($vmid, $resp) = @_;
3116 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3117 $qmpclient->queue_cmd($vmid, $machinecb, 'query-machines');
3118 $qmpclient->queue_cmd($vmid, $versioncb, 'query-version');
3119 # this fails if ballon driver is not loaded, so this must be
3120 # the last commnand (following command are aborted if this fails).
3121 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3123 my $status = 'unknown';
3124 if (!defined($status = $resp->{'return'}->{status
})) {
3125 warn "unable to get VM status\n";
3129 $res->{$vmid}->{qmpstatus
} = $resp->{'return'}->{status
};
3132 foreach my $vmid (keys %$list) {
3133 next if $opt_vmid && ($vmid ne $opt_vmid);
3134 next if !$res->{$vmid}->{pid
}; # not running
3135 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3138 $qmpclient->queue_execute(undef, 2);
3140 foreach my $vmid (keys %$list) {
3141 next if $opt_vmid && ($vmid ne $opt_vmid);
3142 next if !$res->{$vmid}->{pid
}; #not running
3144 # we can't use the $qmpclient since it might have already aborted on
3145 # 'query-balloon', but this might also fail for older versions...
3146 my $qemu_support = eval { mon_cmd
($vmid, "query-proxmox-support") };
3147 $res->{$vmid}->{'proxmox-support'} = $qemu_support // {};
3150 foreach my $vmid (keys %$list) {
3151 next if $opt_vmid && ($vmid ne $opt_vmid);
3152 $res->{$vmid}->{qmpstatus
} = $res->{$vmid}->{status
} if !$res->{$vmid}->{qmpstatus
};
3158 sub conf_has_serial
{
3161 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3162 if ($conf->{"serial$i"}) {
3170 sub conf_has_audio
{
3171 my ($conf, $id) = @_;
3174 my $audio = $conf->{"audio$id"};
3175 return if !defined($audio);
3177 my $audioproperties = parse_property_string
($audio_fmt, $audio);
3178 my $audiodriver = $audioproperties->{driver
} // 'spice';
3181 dev
=> $audioproperties->{device
},
3182 dev_id
=> "audiodev$id",
3183 backend
=> $audiodriver,
3184 backend_id
=> "$audiodriver-backend${id}",
3189 my ($audio, $audiopciaddr, $machine_version) = @_;
3193 my $id = $audio->{dev_id
};
3195 if (min_version
($machine_version, 4, 2)) {
3196 $audiodev = ",audiodev=$audio->{backend_id}";
3199 if ($audio->{dev
} eq 'AC97') {
3200 push @$devs, '-device', "AC97,id=${id}${audiopciaddr}$audiodev";
3201 } elsif ($audio->{dev
} =~ /intel\-hda$/) {
3202 push @$devs, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3203 push @$devs, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0$audiodev";
3204 push @$devs, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1$audiodev";
3206 die "unkown audio device '$audio->{dev}', implement me!";
3209 push @$devs, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3217 socket => "/var/run/qemu-server/$vmid.swtpm",
3218 pid
=> "/var/run/qemu-server/$vmid.swtpm.pid",
3222 sub add_tpm_device
{
3223 my ($vmid, $devices, $conf) = @_;
3225 return if !$conf->{tpmstate0
};
3227 my $paths = get_tpm_paths
($vmid);
3229 push @$devices, "-chardev", "socket,id=tpmchar,path=$paths->{socket}";
3230 push @$devices, "-tpmdev", "emulator,id=tpmdev,chardev=tpmchar";
3231 push @$devices, "-device", "tpm-tis,tpmdev=tpmdev";
3235 my ($storecfg, $vmid, $tpmdrive, $migration) = @_;
3237 return if !$tpmdrive;
3240 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
3241 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
3243 $state = PVE
::Storage
::map_volume
($storecfg, $tpm->{file
});
3245 $state = $tpm->{file
};
3248 my $paths = get_tpm_paths
($vmid);
3250 # during migration, we will get state from remote
3253 # run swtpm_setup to create a new TPM state if it doesn't exist yet
3260 "--create-platform-cert",
3263 "/etc/swtpm_setup.conf", # do not use XDG configs
3265 "0", # force creation as root, error if not possible
3266 "--not-overwrite", # ignore existing state, do not modify
3269 push @$setup_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3270 # TPM 2.0 supports ECC crypto, use if possible
3271 push @$setup_cmd, "--ecc" if $tpm->{version
} eq 'v2.0';
3273 run_command
($setup_cmd, outfunc
=> sub {
3274 print "swtpm_setup: $1\n";
3278 # Used to distinguish different invocations in the log.
3279 my $log_prefix = "[id=" . int(time()) . "] ";
3281 my $emulator_cmd = [
3285 "backend-uri=file://$state,mode=0600",
3287 "type=unixio,path=$paths->{socket},mode=0600",
3289 "file=$paths->{pid}",
3290 "--terminate", # terminate on QEMU disconnect
3293 "file=/run/qemu-server/$vmid-swtpm.log,level=1,prefix=$log_prefix",
3295 push @$emulator_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3296 run_command
($emulator_cmd, outfunc
=> sub { print $1; });
3298 my $tries = 100; # swtpm may take a bit to start before daemonizing, wait up to 5s for pid
3299 while (! -e
$paths->{pid
}) {
3300 die "failed to start swtpm: pid file '$paths->{pid}' wasn't created.\n" if --$tries == 0;
3304 # return untainted PID of swtpm daemon so it can be killed on error
3305 file_read_firstline
($paths->{pid
}) =~ m/(\d+)/;
3309 sub vga_conf_has_spice
{
3312 my $vgaconf = parse_vga
($vga);
3313 my $vgatype = $vgaconf->{type
};
3314 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3321 return get_host_arch
() eq $arch;
3326 return $conf->{arch
} // get_host_arch
();
3329 my $default_machines = {
3334 sub get_installed_machine_version
{
3335 my ($kvmversion) = @_;
3336 $kvmversion = kvm_user_version
() if !defined($kvmversion);
3337 $kvmversion =~ m/^(\d+\.\d+)/;
3341 sub windows_get_pinned_machine_version
{
3342 my ($machine, $base_version, $kvmversion) = @_;
3344 my $pin_version = $base_version;
3345 if (!defined($base_version) ||
3346 !PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($base_version, $kvmversion)
3348 $pin_version = get_installed_machine_version
($kvmversion);
3350 if (!$machine || $machine eq 'pc') {
3351 $machine = "pc-i440fx-$pin_version";
3352 } elsif ($machine eq 'q35') {
3353 $machine = "pc-q35-$pin_version";
3354 } elsif ($machine eq 'virt') {
3355 $machine = "virt-$pin_version";
3357 warn "unknown machine type '$machine', not touching that!\n";
3363 sub get_vm_machine
{
3364 my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
3366 my $machine = $forcemachine || $conf->{machine
};
3368 if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
3369 $kvmversion //= kvm_user_version
();
3370 # we must pin Windows VMs without a specific version to 5.1, as 5.2 fixed a bug in ACPI
3371 # layout which confuses windows quite a bit and may result in various regressions..
3372 # see: https://lists.gnu.org/archive/html/qemu-devel/2021-02/msg08484.html
3373 if (windows_version
($conf->{ostype
})) {
3374 $machine = windows_get_pinned_machine_version
($machine, '5.1', $kvmversion);
3377 $machine ||= $default_machines->{$arch};
3378 if ($add_pve_version) {
3379 my $pvever = PVE
::QemuServer
::Machine
::get_pve_version
($kvmversion);
3380 $machine .= "+pve$pvever";
3384 if ($add_pve_version && $machine !~ m/\+pve\d+?(?:\.pxe)?$/) {
3385 my $is_pxe = $machine =~ m/^(.*?)\.pxe$/;
3386 $machine = $1 if $is_pxe;
3388 # for version-pinned machines that do not include a pve-version (e.g.
3389 # pc-q35-4.1), we assume 0 to keep them stable in case we bump
3390 $machine .= '+pve0';
3392 $machine .= '.pxe' if $is_pxe;
3398 sub get_ovmf_files
($$$) {
3399 my ($arch, $efidisk, $smm) = @_;
3401 my $types = $OVMF->{$arch}
3402 or die "no OVMF images known for architecture '$arch'\n";
3404 my $type = 'default';
3405 if ($arch ne "aarch64" && defined($efidisk->{efitype
}) && $efidisk->{efitype
} eq '4m') {
3406 $type = $smm ?
"4m" : "4m-no-smm";
3407 $type .= '-ms' if $efidisk->{'pre-enrolled-keys'};
3410 my ($ovmf_code, $ovmf_vars) = $types->{$type}->@*;
3411 die "EFI base image '$ovmf_code' not found\n" if ! -f
$ovmf_code;
3412 die "EFI vars image '$ovmf_vars' not found\n" if ! -f
$ovmf_vars;
3414 return ($ovmf_code, $ovmf_vars);
3418 aarch64
=> '/usr/bin/qemu-system-aarch64',
3419 x86_64
=> '/usr/bin/qemu-system-x86_64',
3421 sub get_command_for_arch
($) {
3423 return '/usr/bin/kvm' if is_native
($arch);
3425 my $cmd = $Arch2Qemu->{$arch}
3426 or die "don't know how to emulate architecture '$arch'\n";
3430 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
3431 # to use in a QEMU command line (-cpu element), first array_intersect the result
3432 # of query_supported_ with query_understood_. This is necessary because:
3434 # a) query_understood_ returns flags the host cannot use and
3435 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
3436 # flags, but CPU settings - with most of them being flags. Those settings
3437 # (and some flags, curiously) cannot be specified as a "-cpu" argument.
3439 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
3440 # expensive. If you need the value returned from this, you can get it much
3441 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
3442 # $accel being 'kvm' or 'tcg'.
3444 # pvestatd calls this function on startup and whenever the QEMU/KVM version
3445 # changes, automatically populating pmxcfs.
3447 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
3448 # since kvm and tcg machines support different flags
3450 sub query_supported_cpu_flags
{
3453 $arch //= get_host_arch
();
3454 my $default_machine = $default_machines->{$arch};
3458 # FIXME: Once this is merged, the code below should work for ARM as well:
3459 # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
3460 die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
3463 my $kvm_supported = defined(kvm_version
());
3464 my $qemu_cmd = get_command_for_arch
($arch);
3466 my $pidfile = PVE
::QemuServer
::Helpers
::pidfile_name
($fakevmid);
3468 # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
3469 my $query_supported_run_qemu = sub {
3475 '-machine', $default_machine,
3477 '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server=on,wait=off",
3478 '-mon', 'chardev=qmp,mode=control',
3479 '-pidfile', $pidfile,
3484 push @$cmd, '-accel', 'tcg';
3487 my $rc = run_command
($cmd, noerr
=> 1, quiet
=> 0);
3488 die "QEMU flag querying VM exited with code " . $rc if $rc;
3491 my $cmd_result = mon_cmd
(
3493 'query-cpu-model-expansion',
3495 model
=> { name
=> 'host' }
3498 my $props = $cmd_result->{model
}->{props
};
3499 foreach my $prop (keys %$props) {
3500 next if $props->{$prop} ne '1';
3501 # QEMU returns some flags multiple times, with '_', '.' or '-'
3502 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
3503 # We only keep those with underscores, to match /proc/cpuinfo
3504 $prop =~ s/\.|-/_/g;
3505 $flags->{$prop} = 1;
3510 # force stop with 10 sec timeout and 'nocheck', always stop, even if QMP failed
3511 vm_stop
(undef, $fakevmid, 1, 1, 10, 0, 1);
3515 return [ sort keys %$flags ];
3518 # We need to query QEMU twice, since KVM and TCG have different supported flags
3519 PVE
::QemuConfig-
>lock_config($fakevmid, sub {
3520 $flags->{tcg
} = eval { $query_supported_run_qemu->(0) };
3521 warn "warning: failed querying supported tcg flags: $@\n" if $@;
3523 if ($kvm_supported) {
3524 $flags->{kvm
} = eval { $query_supported_run_qemu->(1) };
3525 warn "warning: failed querying supported kvm flags: $@\n" if $@;
3532 # Understood CPU flags are written to a file at 'pve-qemu' compile time
3533 my $understood_cpu_flag_dir = "/usr/share/kvm";
3534 sub query_understood_cpu_flags
{
3535 my $arch = get_host_arch
();
3536 my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
3538 die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
3541 my $raw = file_get_contents
($filepath);
3542 $raw =~ s/^\s+|\s+$//g;
3543 my @flags = split(/\s+/, $raw);
3548 # Since commit 277d33454f77ec1d1e0bc04e37621e4dd2424b67 in pve-qemu, smm is not off by default
3549 # anymore. But smm=off seems to be required when using SeaBIOS and serial display.
3550 my sub should_disable_smm
{
3551 my ($conf, $vga, $machine) = @_;
3553 return if $machine =~ m/^virt/; # there is no smm flag that could be disabled
3555 return (!defined($conf->{bios
}) || $conf->{bios
} eq 'seabios') &&
3556 $vga->{type
} && $vga->{type
} =~ m/^(serial\d+|none)$/;
3559 my sub print_ovmf_drive_commandlines
{
3560 my ($conf, $storecfg, $vmid, $arch, $q35, $version_guard) = @_;
3562 my $d = $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
3564 my ($ovmf_code, $ovmf_vars) = get_ovmf_files
($arch, $d, $q35);
3566 my $var_drive_str = "if=pflash,unit=1,id=drive-efidisk0";
3568 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($d->{file
}, 1);
3569 my ($path, $format) = $d->@{'file', 'format'};
3571 $path = PVE
::Storage
::path
($storecfg, $d->{file
});
3572 if (!defined($format)) {
3573 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
3574 $format = qemu_img_format
($scfg, $volname);
3576 } elsif (!defined($format)) {
3577 die "efidisk format must be specified\n";
3579 # SPI flash does lots of read-modify-write OPs, without writeback this gets really slow #3329
3580 if ($path =~ m/^rbd:/) {
3581 $var_drive_str .= ',cache=writeback';
3582 $path .= ':rbd_cache_policy=writeback'; # avoid write-around, we *need* to cache writes too
3584 $var_drive_str .= ",format=$format,file=$path";
3586 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $format eq 'raw' && $version_guard->(4, 1, 2);
3587 $var_drive_str .= ',readonly=on' if drive_is_read_only
($conf, $d);
3589 log_warn
("no efidisk configured! Using temporary efivars disk.");
3590 my $path = "/tmp/$vmid-ovmf.fd";
3591 PVE
::Tools
::file_copy
($ovmf_vars, $path, -s
$ovmf_vars);
3592 $var_drive_str .= ",format=raw,file=$path";
3593 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $version_guard->(4, 1, 2);
3596 return ("if=pflash,unit=0,format=raw,readonly=on,file=$ovmf_code", $var_drive_str);
3599 sub config_to_command
{
3600 my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu,
3603 my ($globalFlags, $machineFlags, $rtcFlags) = ([], [], []);
3606 my $ostype = $conf->{ostype
};
3607 my $winversion = windows_version
($ostype);
3608 my $kvm = $conf->{kvm
};
3609 my $nodename = nodename
();
3611 my $arch = get_vm_arch
($conf);
3612 my $kvm_binary = get_command_for_arch
($arch);
3613 my $kvmver = kvm_user_version
($kvm_binary);
3615 if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
3616 $kvmver //= "undefined";
3617 die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
3620 my $add_pve_version = min_version
($kvmver, 4, 1);
3622 my $machine_type = get_vm_machine
($conf, $forcemachine, $arch, $add_pve_version);
3623 my $machine_version = extract_version
($machine_type, $kvmver);
3624 $kvm //= 1 if is_native
($arch);
3626 $machine_version =~ m/(\d+)\.(\d+)/;
3627 my ($machine_major, $machine_minor) = ($1, $2);
3629 if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
3630 warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
3631 } elsif (!min_version
($kvmver, $machine_major, $machine_minor)) {
3632 die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type',"
3633 ." please upgrade node '$nodename'\n"
3634 } elsif (!PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($machine_version, $kvmver)) {
3635 my $max_pve_version = PVE
::QemuServer
::Machine
::get_pve_version
($machine_version);
3636 die "Installed qemu-server (max feature level for $machine_major.$machine_minor is"
3637 ." pve$max_pve_version) is too old to run machine type '$machine_type', please upgrade"
3638 ." node '$nodename'\n";
3641 # if a specific +pve version is required for a feature, use $version_guard
3642 # instead of min_version to allow machines to be run with the minimum
3644 my $required_pve_version = 0;
3645 my $version_guard = sub {
3646 my ($major, $minor, $pve) = @_;
3647 return 0 if !min_version
($machine_version, $major, $minor, $pve);
3648 my $max_pve = PVE
::QemuServer
::Machine
::get_pve_version
("$major.$minor");
3649 return 1 if min_version
($machine_version, $major, $minor, $max_pve+1);
3650 $required_pve_version = $pve if $pve && $pve > $required_pve_version;
3654 if ($kvm && !defined kvm_version
()) {
3655 die "KVM virtualisation configured, but not available. Either disable in VM configuration"
3656 ." or enable in BIOS.\n";
3659 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
3660 my $hotplug_features = parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
3661 my $use_old_bios_files = undef;
3662 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
3665 if ($conf->{affinity
}) {
3666 push @$cmd, '/usr/bin/taskset', '--cpu-list', '--all-tasks', $conf->{affinity
};
3669 push @$cmd, $kvm_binary;
3671 push @$cmd, '-id', $vmid;
3673 my $vmname = $conf->{name
} || "vm$vmid";
3675 push @$cmd, '-name', "$vmname,debug-threads=on";
3677 push @$cmd, '-no-shutdown';
3681 my $qmpsocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid);
3682 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server=on,wait=off";
3683 push @$cmd, '-mon', "chardev=qmp,mode=control";
3685 if (min_version
($machine_version, 2, 12)) {
3686 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3687 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3690 push @$cmd, '-pidfile' , PVE
::QemuServer
::Helpers
::pidfile_name
($vmid);
3692 push @$cmd, '-daemonize';
3694 if ($conf->{smbios1
}) {
3695 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
3696 if ($smbios_conf->{base64
}) {
3697 # Do not pass base64 flag to qemu
3698 delete $smbios_conf->{base64
};
3699 my $smbios_string = "";
3700 foreach my $key (keys %$smbios_conf) {
3702 if ($key eq "uuid") {
3703 $value = $smbios_conf->{uuid
}
3705 $value = decode_base64
($smbios_conf->{$key});
3707 # qemu accepts any binary data, only commas need escaping by double comma
3709 $smbios_string .= "," . $key . "=" . $value if $value;
3711 push @$cmd, '-smbios', "type=1" . $smbios_string;
3713 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3717 if ($conf->{bios
} && $conf->{bios
} eq 'ovmf') {
3718 my ($code_drive_str, $var_drive_str) =
3719 print_ovmf_drive_commandlines
($conf, $storecfg, $vmid, $arch, $q35, $version_guard);
3720 push $cmd->@*, '-drive', $code_drive_str;
3721 push $cmd->@*, '-drive', $var_drive_str;
3724 if ($q35) { # tell QEMU to load q35 config early
3725 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3726 if (min_version
($machine_version, 4, 0)) {
3727 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3729 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3733 if (defined(my $fixups = qemu_created_version_fixups
($conf, $forcemachine, $kvmver))) {
3734 push @$cmd, $fixups->@*;
3737 if ($conf->{vmgenid
}) {
3738 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid
};
3741 # add usb controllers
3742 my @usbcontrollers = PVE
::QemuServer
::USB
::get_usb_controllers
(
3743 $conf, $bridges, $arch, $machine_type, $usbdesc->{format
}, $MAX_USB_DEVICES, $machine_version);
3744 push @$devices, @usbcontrollers if @usbcontrollers;
3745 my $vga = parse_vga
($conf->{vga
});
3747 my $qxlnum = vga_conf_has_spice
($conf->{vga
});
3748 $vga->{type
} = 'qxl' if $qxlnum;
3750 if (!$vga->{type
}) {
3751 if ($arch eq 'aarch64') {
3752 $vga->{type
} = 'virtio';
3753 } elsif (min_version
($machine_version, 2, 9)) {
3754 $vga->{type
} = (!$winversion || $winversion >= 6) ?
'std' : 'cirrus';
3756 $vga->{type
} = ($winversion >= 6) ?
'std' : 'cirrus';
3760 # enable absolute mouse coordinates (needed by vnc)
3761 my $tablet = $conf->{tablet
};
3762 if (!defined($tablet)) {
3763 $tablet = $defaults->{tablet
};
3764 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3765 $tablet = 0 if $vga->{type
} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3769 push @$devices, '-device', print_tabletdevice_full
($conf, $arch) if $tablet;
3770 my $kbd = print_keyboarddevice_full
($conf, $arch);
3771 push @$devices, '-device', $kbd if defined($kbd);
3774 my $bootorder = device_bootorder
($conf);
3776 # host pci device passthrough
3777 my ($kvm_off, $gpu_passthrough, $legacy_igd) = PVE
::QemuServer
::PCI
::print_hostpci_devices
(
3778 $vmid, $conf, $devices, $vga, $winversion, $q35, $bridges, $arch, $machine_type, $bootorder);
3781 my $usb_dev_features = {};
3782 $usb_dev_features->{spice_usb3
} = 1 if min_version
($machine_version, 4, 0);
3784 my @usbdevices = PVE
::QemuServer
::USB
::get_usb_devices
(
3785 $conf, $usbdesc->{format
}, $MAX_USB_DEVICES, $usb_dev_features, $bootorder, $machine_version);
3786 push @$devices, @usbdevices if @usbdevices;
3789 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3790 my $path = $conf->{"serial$i"} or next;
3791 if ($path eq 'socket') {
3792 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3793 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server=on,wait=off";
3794 # On aarch64, serial0 is the UART device. QEMU only allows
3795 # connecting UART devices via the '-serial' command line, as
3796 # the device has a fixed slot on the hardware...
3797 if ($arch eq 'aarch64' && $i == 0) {
3798 push @$devices, '-serial', "chardev:serial$i";
3800 push @$devices, '-device', "isa-serial,chardev=serial$i";
3803 die "no such serial device\n" if ! -c
$path;
3804 push @$devices, '-chardev', "serial,id=serial$i,path=$path";
3805 push @$devices, '-device', "isa-serial,chardev=serial$i";
3810 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3811 if (my $path = $conf->{"parallel$i"}) {
3812 die "no such parallel device\n" if ! -c
$path;
3813 my $devtype = $path =~ m!^/dev/usb/lp! ?
'serial' : 'parallel';
3814 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3815 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3819 if (min_version
($machine_version, 4, 0) && (my $audio = conf_has_audio
($conf))) {
3820 my $audiopciaddr = print_pci_addr
("audio0", $bridges, $arch, $machine_type);
3821 my $audio_devs = audio_devs
($audio, $audiopciaddr, $machine_version);
3822 push @$devices, @$audio_devs;
3825 add_tpm_device
($vmid, $devices, $conf);
3828 $sockets = $conf->{smp
} if $conf->{smp
}; # old style - no longer iused
3829 $sockets = $conf->{sockets
} if $conf->{sockets
};
3831 my $cores = $conf->{cores
} || 1;
3833 my $maxcpus = $sockets * $cores;
3835 my $vcpus = $conf->{vcpus
} ?
$conf->{vcpus
} : $maxcpus;
3837 my $allowed_vcpus = $cpuinfo->{cpus
};
3839 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n" if ($allowed_vcpus < $maxcpus);
3841 if ($hotplug_features->{cpu
} && min_version
($machine_version, 2, 7)) {
3842 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3843 for (my $i = 2; $i <= $vcpus; $i++) {
3844 my $cpustr = print_cpu_device
($conf,$i);
3845 push @$cmd, '-device', $cpustr;
3850 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3852 push @$cmd, '-nodefaults';
3854 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3856 push $machineFlags->@*, 'acpi=off' if defined($conf->{acpi
}) && $conf->{acpi
} == 0;
3858 push @$cmd, '-no-reboot' if defined($conf->{reboot
}) && $conf->{reboot
} == 0;
3860 if ($vga->{type
} && $vga->{type
} !~ m/^serial\d+$/ && $vga->{type
} ne 'none'){
3861 push @$devices, '-device', print_vga_device
(
3862 $conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3864 push @$cmd, '-display', 'egl-headless,gl=core' if $vga->{type
} eq 'virtio-gl'; # VIRGL
3866 my $socket = PVE
::QemuServer
::Helpers
::vnc_socket
($vmid);
3867 push @$cmd, '-vnc', "unix:$socket,password=on";
3869 push @$cmd, '-vga', 'none' if $vga->{type
} eq 'none';
3870 push @$cmd, '-nographic';
3874 my $tdf = defined($conf->{tdf
}) ?
$conf->{tdf
} : $defaults->{tdf
};
3875 my $useLocaltime = $conf->{localtime};
3877 if ($winversion >= 5) { # windows
3878 $useLocaltime = 1 if !defined($conf->{localtime});
3880 # use time drift fix when acpi is enabled
3881 if (!(defined($conf->{acpi
}) && $conf->{acpi
} == 0)) {
3882 $tdf = 1 if !defined($conf->{tdf
});
3886 if ($winversion >= 6) {
3887 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3888 push @$machineFlags, 'hpet=off';
3891 push @$rtcFlags, 'driftfix=slew' if $tdf;
3893 if ($conf->{startdate
} && $conf->{startdate
} ne 'now') {
3894 push @$rtcFlags, "base=$conf->{startdate}";
3895 } elsif ($useLocaltime) {
3896 push @$rtcFlags, 'base=localtime';
3900 push @$cmd, '-cpu', $forcecpu;
3902 push @$cmd, get_cpu_options
($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough);
3905 PVE
::QemuServer
::Memory
::config
(
3906 $conf, $vmid, $sockets, $cores, $defaults, $hotplug_features->{memory
}, $cmd);
3908 push @$cmd, '-S' if $conf->{freeze
};
3910 push @$cmd, '-k', $conf->{keyboard
} if defined($conf->{keyboard
});
3912 my $guest_agent = parse_guest_agent
($conf);
3914 if ($guest_agent->{enabled
}) {
3915 my $qgasocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid, 1);
3916 push @$devices, '-chardev', "socket,path=$qgasocket,server=on,wait=off,id=qga0";
3918 if (!$guest_agent->{type
} || $guest_agent->{type
} eq 'virtio') {
3919 my $pciaddr = print_pci_addr
("qga0", $bridges, $arch, $machine_type);
3920 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3921 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3922 } elsif ($guest_agent->{type
} eq 'isa') {
3923 push @$devices, '-device', "isa-serial,chardev=qga0";
3927 my $rng = $conf->{rng0
} ? parse_rng
($conf->{rng0
}) : undef;
3928 if ($rng && $version_guard->(4, 1, 2)) {
3929 check_rng_source
($rng->{source
});
3931 my $max_bytes = $rng->{max_bytes
} // $rng_fmt->{max_bytes
}->{default};
3932 my $period = $rng->{period
} // $rng_fmt->{period
}->{default};
3933 my $limiter_str = "";
3935 $limiter_str = ",max-bytes=$max_bytes,period=$period";
3938 my $rng_addr = print_pci_addr
("rng0", $bridges, $arch, $machine_type);
3939 push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3940 push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3945 if ($qxlnum || $vga->{type
} =~ /^virtio/) {
3948 for (my $i = 1; $i < $qxlnum; $i++){
3949 push @$devices, '-device', print_vga_device
(
3950 $conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3953 # assume other OS works like Linux
3954 my ($ram, $vram) = ("134217728", "67108864");
3955 if ($vga->{memory
}) {
3956 $ram = PVE
::Tools
::convert_size
($qxlnum*4*$vga->{memory
}, 'mb' => 'b');
3957 $vram = PVE
::Tools
::convert_size
($qxlnum*2*$vga->{memory
}, 'mb' => 'b');
3959 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3960 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3964 my $pciaddr = print_pci_addr
("spice", $bridges, $arch, $machine_type);
3966 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
3967 my @nodeaddrs = PVE
::Tools
::getaddrinfo_all
('localhost', family
=> $pfamily);
3968 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3970 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3971 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3972 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3974 my $localhost = PVE
::Network
::addr_to_ip
($nodeaddrs[0]->{addr
});
3975 $spice_port = PVE
::Tools
::next_spice_port
($pfamily, $localhost);
3977 my $spice_enhancement_str = $conf->{spice_enhancements
} // '';
3978 my $spice_enhancement = parse_property_string
($spice_enhancements_fmt, $spice_enhancement_str);
3979 if ($spice_enhancement->{foldersharing
}) {
3980 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3981 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3984 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3985 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}"
3986 if $spice_enhancement->{videostreaming
};
3988 push @$devices, '-spice', "$spice_opts";
3991 # enable balloon by default, unless explicitly disabled
3992 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
3993 my $pciaddr = print_pci_addr
("balloon0", $bridges, $arch, $machine_type);
3994 my $ballooncmd = "virtio-balloon-pci,id=balloon0$pciaddr";
3995 $ballooncmd .= ",free-page-reporting=on" if min_version
($machine_version, 6, 2);
3996 push @$devices, '-device', $ballooncmd;
3999 if ($conf->{watchdog
}) {
4000 my $wdopts = parse_watchdog
($conf->{watchdog
});
4001 my $pciaddr = print_pci_addr
("watchdog", $bridges, $arch, $machine_type);
4002 my $watchdog = $wdopts->{model
} || 'i6300esb';
4003 push @$devices, '-device', "$watchdog$pciaddr";
4004 push @$devices, '-watchdog-action', $wdopts->{action
} if $wdopts->{action
};
4008 my $scsicontroller = {};
4009 my $ahcicontroller = {};
4010 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : $defaults->{scsihw
};
4012 # Add iscsi initiator name if available
4013 if (my $initiator = get_initiator_name
()) {
4014 push @$devices, '-iscsi', "initiator-name=$initiator";
4017 PVE
::QemuConfig-
>foreach_volume($conf, sub {
4018 my ($ds, $drive) = @_;
4020 if (PVE
::Storage
::parse_volume_id
($drive->{file
}, 1)) {
4021 check_volume_storage_type
($storecfg, $drive->{file
});
4022 push @$vollist, $drive->{file
};
4025 # ignore efidisk here, already added in bios/fw handling code above
4026 return if $drive->{interface
} eq 'efidisk';
4028 return if $drive->{interface
} eq 'tpmstate';
4030 $use_virtio = 1 if $ds =~ m/^virtio/;
4032 $drive->{bootindex
} = $bootorder->{$ds} if $bootorder->{$ds};
4034 if ($drive->{interface
} eq 'virtio'){
4035 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread
};
4038 if ($drive->{interface
} eq 'scsi') {
4040 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
4042 die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
4043 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
4045 my $pciaddr = print_pci_addr
("$controller_prefix$controller", $bridges, $arch, $machine_type);
4046 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ?
"virtio-scsi-pci" : $scsihw;
4049 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{iothread
}){
4050 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4051 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4052 } elsif ($drive->{iothread
}) {
4054 "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n"
4059 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{queues
}){
4060 $queues = ",num_queues=$drive->{queues}";
4063 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues"
4064 if !$scsicontroller->{$controller};
4065 $scsicontroller->{$controller}=1;
4068 if ($drive->{interface
} eq 'sata') {
4069 my $controller = int($drive->{index} / $PVE::QemuServer
::Drive
::MAX_SATA_DISKS
);
4070 my $pciaddr = print_pci_addr
("ahci$controller", $bridges, $arch, $machine_type);
4071 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr"
4072 if !$ahcicontroller->{$controller};
4073 $ahcicontroller->{$controller}=1;
4076 my $pbs_conf = $pbs_backing->{$ds};
4077 my $pbs_name = undef;
4079 $pbs_name = "drive-$ds-pbs";
4080 push @$devices, '-blockdev', print_pbs_blockdev
($pbs_conf, $pbs_name);
4083 my $drive_cmd = print_drive_commandline_full
(
4084 $storecfg, $vmid, $drive, $pbs_name, min_version
($kvmver, 6, 0));
4086 # extra protection for templates, but SATA and IDE don't support it..
4087 $drive_cmd .= ',readonly=on' if drive_is_read_only
($conf, $drive);
4089 push @$devices, '-drive',$drive_cmd;
4090 push @$devices, '-device', print_drivedevice_full
(
4091 $storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4094 for (my $i = 0; $i < $MAX_NETS; $i++) {
4095 my $netname = "net$i";
4097 next if !$conf->{$netname};
4098 my $d = parse_net
($conf->{$netname});
4100 # save the MAC addr here (could be auto-gen. in some odd setups) for FDB registering later?
4102 $use_virtio = 1 if $d->{model
} eq 'virtio';
4104 $d->{bootindex
} = $bootorder->{$netname} if $bootorder->{$netname};
4106 my $netdevfull = print_netdev_full
($vmid, $conf, $arch, $d, $netname);
4107 push @$devices, '-netdev', $netdevfull;
4109 my $netdevicefull = print_netdevice_full
(
4110 $vmid, $conf, $d, $netname, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version);
4112 push @$devices, '-device', $netdevicefull;
4115 if ($conf->{ivshmem
}) {
4116 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
4120 $bus = print_pcie_addr
("ivshmem");
4122 $bus = print_pci_addr
("ivshmem", $bridges, $arch, $machine_type);
4125 my $ivshmem_name = $ivshmem->{name
} // $vmid;
4126 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4128 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4129 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path"
4130 .",size=$ivshmem->{size}M";
4133 # pci.4 is nested in pci.1
4134 $bridges->{1} = 1 if $bridges->{4};
4136 if (!$q35) { # add pci bridges
4137 if (min_version
($machine_version, 2, 3)) {
4141 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4144 for my $k (sort {$b cmp $a} keys %$bridges) {
4145 next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
4148 if ($k == 2 && $legacy_igd) {
4151 my $pciaddr = print_pci_addr
("pci.$k_name", undef, $arch, $machine_type);
4152 my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
4154 if ($q35) { # add after -readconfig pve-q35.cfg
4155 splice @$devices, 2, 0, '-device', $devstr;
4157 unshift @$devices, '-device', $devstr if $k > 0;
4162 push @$machineFlags, 'accel=tcg';
4165 push @$machineFlags, 'smm=off' if should_disable_smm
($conf, $vga, $machine_type);
4167 my $machine_type_min = $machine_type;
4168 if ($add_pve_version) {
4169 $machine_type_min =~ s/\+pve\d+$//;
4170 $machine_type_min .= "+pve$required_pve_version";
4172 push @$machineFlags, "type=${machine_type_min}";
4174 push @$cmd, @$devices;
4175 push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
4176 push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
4177 push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
4179 if (my $vmstate = $conf->{vmstate
}) {
4180 my $statepath = PVE
::Storage
::path
($storecfg, $vmstate);
4181 push @$vollist, $vmstate;
4182 push @$cmd, '-loadstate', $statepath;
4183 print "activating and using '$vmstate' as vmstate\n";
4186 if (PVE
::QemuConfig-
>is_template($conf)) {
4187 # needed to workaround base volumes being read-only
4188 push @$cmd, '-snapshot';
4192 if ($conf->{args
}) {
4193 my $aa = PVE
::Tools
::split_args
($conf->{args
});
4197 return wantarray ?
($cmd, $vollist, $spice_port) : $cmd;
4200 sub check_rng_source
{
4203 # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
4204 die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
4207 my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
4208 if ($source eq '/dev/hwrng' && file_read_firstline
($rng_current) eq 'none') {
4209 # Needs to abort, otherwise QEMU crashes on first rng access. Note that rng_current cannot
4210 # be changed to 'none' manually, so once the VM is past this point, it's no longer an issue.
4211 die "Cannot start VM with passed-through RNG device: '/dev/hwrng' exists, but"
4212 ." '$rng_current' is set to 'none'. Ensure that a compatible hardware-RNG is attached"
4220 my $res = mon_cmd
($vmid, 'query-spice');
4222 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4225 sub vm_devices_list
{
4228 my $res = mon_cmd
($vmid, 'query-pci');
4229 my $devices_to_check = [];
4231 foreach my $pcibus (@$res) {
4232 push @$devices_to_check, @{$pcibus->{devices
}},
4235 while (@$devices_to_check) {
4237 for my $d (@$devices_to_check) {
4238 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4239 next if !$d->{'pci_bridge'} || !$d->{'pci_bridge'}->{devices
};
4241 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices
}});
4242 push @$to_check, @{$d->{'pci_bridge'}->{devices
}};
4244 $devices_to_check = $to_check;
4247 my $resblock = mon_cmd
($vmid, 'query-block');
4248 foreach my $block (@$resblock) {
4249 if($block->{device
} =~ m/^drive-(\S+)/){
4254 my $resmice = mon_cmd
($vmid, 'query-mice');
4255 foreach my $mice (@$resmice) {
4256 if ($mice->{name
} eq 'QEMU HID Tablet') {
4257 $devices->{tablet
} = 1;
4262 # for usb devices there is no query-usb
4263 # but we can iterate over the entries in
4264 # qom-list path=/machine/peripheral
4265 my $resperipheral = mon_cmd
($vmid, 'qom-list', path
=> '/machine/peripheral');
4266 foreach my $per (@$resperipheral) {
4267 if ($per->{name
} =~ m/^usb(?:redirdev)?\d+$/) {
4268 $devices->{$per->{name
}} = 1;
4276 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4278 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
4280 my $devices_list = vm_devices_list
($vmid);
4281 return 1 if defined($devices_list->{$deviceid});
4283 # add PCI bridge if we need it for the device
4284 qemu_add_pci_bridge
($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type);
4286 if ($deviceid eq 'tablet') {
4287 qemu_deviceadd
($vmid, print_tabletdevice_full
($conf, $arch));
4288 } elsif ($deviceid eq 'keyboard') {
4289 qemu_deviceadd
($vmid, print_keyboarddevice_full
($conf, $arch));
4290 } elsif ($deviceid =~ m/^usbredirdev(\d+)$/) {
4292 qemu_spice_usbredir_chardev_add
($vmid, "usbredirchardev$id");
4293 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_spice_usbdevice
($id, "xhci", $id + 1));
4294 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4295 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_usbdevice_full
($conf, $deviceid, $device, {}, $1 + 1));
4296 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4297 qemu_iothread_add
($vmid, $deviceid, $device);
4299 qemu_driveadd
($storecfg, $vmid, $device);
4300 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4302 qemu_deviceadd
($vmid, $devicefull);
4303 eval { qemu_deviceaddverify
($vmid, $deviceid); };
4305 eval { qemu_drivedel
($vmid, $deviceid); };
4309 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4310 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : "lsi";
4311 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4312 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ?
"virtio-scsi-pci" : $scsihw;
4314 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4316 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread
}) {
4317 qemu_iothread_add
($vmid, $deviceid, $device);
4318 $devicefull .= ",iothread=iothread-$deviceid";
4321 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues
}) {
4322 $devicefull .= ",num_queues=$device->{queues}";
4325 qemu_deviceadd
($vmid, $devicefull);
4326 qemu_deviceaddverify
($vmid, $deviceid);
4327 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4328 qemu_findorcreatescsihw
($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4329 qemu_driveadd
($storecfg, $vmid, $device);
4331 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4332 eval { qemu_deviceadd
($vmid, $devicefull); };
4334 eval { qemu_drivedel
($vmid, $deviceid); };
4338 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4339 return if !qemu_netdevadd
($vmid, $conf, $arch, $device, $deviceid);
4341 my $machine_type = PVE
::QemuServer
::Machine
::qemu_machine_pxe
($vmid, $conf);
4342 my $machine_version = PVE
::QemuServer
::Machine
::extract_version
($machine_type);
4343 my $use_old_bios_files = undef;
4344 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
4346 my $netdevicefull = print_netdevice_full
(
4347 $vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type, $machine_version);
4348 qemu_deviceadd
($vmid, $netdevicefull);
4350 qemu_deviceaddverify
($vmid, $deviceid);
4351 qemu_set_link_status
($vmid, $deviceid, !$device->{link_down
});
4354 eval { qemu_netdevdel
($vmid, $deviceid); };
4358 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4360 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4361 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4363 qemu_deviceadd
($vmid, $devicefull);
4364 qemu_deviceaddverify
($vmid, $deviceid);
4366 die "can't hotplug device '$deviceid'\n";
4372 # fixme: this should raise exceptions on error!
4373 sub vm_deviceunplug
{
4374 my ($vmid, $conf, $deviceid) = @_;
4376 my $devices_list = vm_devices_list
($vmid);
4377 return 1 if !defined($devices_list->{$deviceid});
4379 my $bootdisks = PVE
::QemuServer
::Drive
::get_bootdisks
($conf);
4380 die "can't unplug bootdisk '$deviceid'\n" if grep {$_ eq $deviceid} @$bootdisks;
4382 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard' || $deviceid eq 'xhci') {
4383 qemu_devicedel
($vmid, $deviceid);
4384 } elsif ($deviceid =~ m/^usbredirdev\d+$/) {
4385 qemu_devicedel
($vmid, $deviceid);
4386 qemu_devicedelverify
($vmid, $deviceid);
4387 } elsif ($deviceid =~ m/^usb\d+$/) {
4388 qemu_devicedel
($vmid, $deviceid);
4389 qemu_devicedelverify
($vmid, $deviceid);
4390 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4391 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4393 qemu_devicedel
($vmid, $deviceid);
4394 qemu_devicedelverify
($vmid, $deviceid);
4395 qemu_drivedel
($vmid, $deviceid);
4396 qemu_iothread_del
($vmid, $deviceid, $device);
4397 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4398 qemu_devicedel
($vmid, $deviceid);
4399 qemu_devicedelverify
($vmid, $deviceid);
4400 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4401 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4403 qemu_devicedel
($vmid, $deviceid);
4404 qemu_devicedelverify
($vmid, $deviceid);
4405 qemu_drivedel
($vmid, $deviceid);
4406 qemu_deletescsihw
($conf, $vmid, $deviceid);
4408 qemu_iothread_del
($vmid, "virtioscsi$device->{index}", $device)
4409 if $conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single');
4410 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4411 qemu_devicedel
($vmid, $deviceid);
4412 qemu_devicedelverify
($vmid, $deviceid);
4413 qemu_netdevdel
($vmid, $deviceid);
4415 die "can't unplug device '$deviceid'\n";
4421 sub qemu_spice_usbredir_chardev_add
{
4422 my ($vmid, $id) = @_;
4424 mon_cmd
($vmid, "chardev-add" , (
4435 sub qemu_deviceadd
{
4436 my ($vmid, $devicefull) = @_;
4438 $devicefull = "driver=".$devicefull;
4439 my %options = split(/[=,]/, $devicefull);
4441 mon_cmd
($vmid, "device_add" , %options);
4444 sub qemu_devicedel
{
4445 my ($vmid, $deviceid) = @_;
4447 my $ret = mon_cmd
($vmid, "device_del", id
=> $deviceid);
4450 sub qemu_iothread_add
{
4451 my ($vmid, $deviceid, $device) = @_;
4453 if ($device->{iothread
}) {
4454 my $iothreads = vm_iothreads_list
($vmid);
4455 qemu_objectadd
($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4459 sub qemu_iothread_del
{
4460 my ($vmid, $deviceid, $device) = @_;
4462 if ($device->{iothread
}) {
4463 my $iothreads = vm_iothreads_list
($vmid);
4464 qemu_objectdel
($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4468 sub qemu_objectadd
{
4469 my ($vmid, $objectid, $qomtype) = @_;
4471 mon_cmd
($vmid, "object-add", id
=> $objectid, "qom-type" => $qomtype);
4476 sub qemu_objectdel
{
4477 my ($vmid, $objectid) = @_;
4479 mon_cmd
($vmid, "object-del", id
=> $objectid);
4485 my ($storecfg, $vmid, $device) = @_;
4487 my $kvmver = get_running_qemu_version
($vmid);
4488 my $io_uring = min_version
($kvmver, 6, 0);
4489 my $drive = print_drive_commandline_full
($storecfg, $vmid, $device, undef, $io_uring);
4490 $drive =~ s/\\/\\\\/g;
4491 my $ret = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "drive_add auto \"$drive\"");
4493 # If the command succeeds qemu prints: "OK
"
4494 return 1 if $ret =~ m/OK/s;
4496 die "adding drive failed
: $ret\n";
4500 my ($vmid, $deviceid) = @_;
4502 my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-
$deviceid");
4505 return 1 if $ret eq "";
4507 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4508 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4510 die "deleting drive
$deviceid failed
: $ret\n";
4513 sub qemu_deviceaddverify {
4514 my ($vmid, $deviceid) = @_;
4516 for (my $i = 0; $i <= 5; $i++) {
4517 my $devices_list = vm_devices_list($vmid);
4518 return 1 if defined($devices_list->{$deviceid});
4522 die "error on hotplug device
'$deviceid'\n";
4526 sub qemu_devicedelverify {
4527 my ($vmid, $deviceid) = @_;
4529 # need to verify that the device is correctly removed as device_del
4530 # is async and empty return is not reliable
4532 for (my $i = 0; $i <= 5; $i++) {
4533 my $devices_list = vm_devices_list($vmid);
4534 return 1 if !defined($devices_list->{$deviceid});
4538 die "error on hot-unplugging device
'$deviceid'\n";
4541 sub qemu_findorcreatescsihw {
4542 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4544 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4546 my $scsihwid="$controller_prefix$controller";
4547 my $devices_list = vm_devices_list($vmid);
4549 if (!defined($devices_list->{$scsihwid})) {
4550 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4556 sub qemu_deletescsihw {
4557 my ($conf, $vmid, $opt) = @_;
4559 my $device = parse_drive($opt, $conf->{$opt});
4561 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4562 vm_deviceunplug($vmid, $conf, "virtioscsi
$device->{index}");
4566 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4568 my $devices_list = vm_devices_list($vmid);
4569 foreach my $opt (keys %{$devices_list}) {
4570 if (is_valid_drivename($opt)) {
4571 my $drive = parse_drive($opt, $conf->{$opt});
4572 if ($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4578 my $scsihwid="scsihw
$controller";
4580 vm_deviceunplug($vmid, $conf, $scsihwid);
4585 sub qemu_add_pci_bridge {
4586 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4592 print_pci_addr($device, $bridges, $arch, $machine_type);
4594 while (my ($k, $v) = each %$bridges) {
4597 return 1 if !defined($bridgeid) || $bridgeid < 1;
4599 my $bridge = "pci
.$bridgeid";
4600 my $devices_list = vm_devices_list($vmid);
4602 if (!defined($devices_list->{$bridge})) {
4603 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4609 sub qemu_set_link_status {
4610 my ($vmid, $device, $up) = @_;
4612 mon_cmd($vmid, "set_link
", name => $device,
4613 up => $up ? JSON::true : JSON::false);
4616 sub qemu_netdevadd {
4617 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4619 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4620 my %options = split(/[=,]/, $netdev);
4622 if (defined(my $vhost = $options{vhost})) {
4623 $options{vhost} = JSON::boolean(PVE::JSONSchema::parse_boolean($vhost));
4626 if (defined(my $queues = $options{queues})) {
4627 $options{queues} = $queues + 0;
4630 mon_cmd($vmid, "netdev_add
", %options);
4634 sub qemu_netdevdel {
4635 my ($vmid, $deviceid) = @_;
4637 mon_cmd($vmid, "netdev_del
", id => $deviceid);
4640 sub qemu_usb_hotplug {
4641 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4645 # remove the old one first
4646 vm_deviceunplug($vmid, $conf, $deviceid);
4648 # check if xhci controller is necessary and available
4649 my $devicelist = vm_devices_list($vmid);
4651 if (!$devicelist->{xhci}) {
4652 my $pciaddr = print_pci_addr("xhci
", undef, $arch, $machine_type);
4653 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_qemu_xhci_controller($pciaddr));
4656 # print_usbdevice_full expects the parsed device
4657 my $d = parse_usb_device($device->{host});
4658 $d->{usb3} = $device->{usb3};
4661 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4664 sub qemu_cpu_hotplug {
4665 my ($vmid, $conf, $vcpus) = @_;
4667 my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4670 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4671 $sockets = $conf->{sockets} if $conf->{sockets};
4672 my $cores = $conf->{cores} || 1;
4673 my $maxcpus = $sockets * $cores;
4675 $vcpus = $maxcpus if !$vcpus;
4677 die "you can
't add more vcpus than maxcpus\n"
4678 if $vcpus > $maxcpus;
4680 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4682 if ($vcpus < $currentvcpus) {
4684 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4686 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4687 qemu_devicedel($vmid, "cpu$i");
4689 my $currentrunningvcpus = undef;
4691 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4692 last if scalar(@{$currentrunningvcpus}) == $i-1;
4693 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4697 #update conf after each succesfull cpu unplug
4698 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4699 PVE::QemuConfig->write_config($vmid, $conf);
4702 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4708 my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4709 die "vcpus in running vm does not match its configuration\n"
4710 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4712 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4714 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4715 my $cpustr = print_cpu_device($conf, $i);
4716 qemu_deviceadd($vmid, $cpustr);
4719 my $currentrunningvcpus = undef;
4721 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4722 last if scalar(@{$currentrunningvcpus}) == $i;
4723 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4727 #update conf after each succesfull cpu hotplug
4728 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4729 PVE::QemuConfig->write_config($vmid, $conf);
4733 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4734 mon_cmd($vmid, "cpu-add", id => int($i));
4739 sub qemu_block_set_io_throttle {
4740 my ($vmid, $deviceid,
4741 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4742 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4743 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4744 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4746 return if !check_running($vmid) ;
4748 mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4750 bps_rd => int($bps_rd),
4751 bps_wr => int($bps_wr),
4753 iops_rd => int($iops_rd),
4754 iops_wr => int($iops_wr),
4755 bps_max => int($bps_max),
4756 bps_rd_max => int($bps_rd_max),
4757 bps_wr_max => int($bps_wr_max),
4758 iops_max => int($iops_max),
4759 iops_rd_max => int($iops_rd_max),
4760 iops_wr_max => int($iops_wr_max),
4761 bps_max_length => int($bps_max_length),
4762 bps_rd_max_length => int($bps_rd_max_length),
4763 bps_wr_max_length => int($bps_wr_max_length),
4764 iops_max_length => int($iops_max_length),
4765 iops_rd_max_length => int($iops_rd_max_length),
4766 iops_wr_max_length => int($iops_wr_max_length),
4771 sub qemu_block_resize {
4772 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4774 my $running = check_running($vmid);
4776 PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4778 return if !$running;
4780 my $padding = (1024 - $size % 1024) % 1024;
4781 $size = $size + $padding;
4786 device => $deviceid,
4792 sub qemu_volume_snapshot {
4793 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4795 my $running = check_running($vmid);
4797 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4798 mon_cmd($vmid, 'blockdev-snapshot-internal-sync
', device => $deviceid, name => $snap);
4800 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4804 sub qemu_volume_snapshot_delete {
4805 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4807 my $running = check_running($vmid);
4812 my $conf = PVE::QemuConfig->load_config($vmid);
4813 PVE::QemuConfig->foreach_volume($conf, sub {
4814 my ($ds, $drive) = @_;
4815 $running = 1 if $drive->{file} eq $volid;
4819 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4820 mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync
', device => $deviceid, name => $snap);
4822 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4826 sub set_migration_caps {
4827 my ($vmid, $savevm) = @_;
4829 my $qemu_support = eval { mon_cmd($vmid, "query-proxmox-support") };
4831 my $bitmap_prop = $savevm ? 'pbs-dirty-bitmap-savevm
' : 'pbs-dirty-bitmap-migration
';
4832 my $dirty_bitmaps = $qemu_support->{$bitmap_prop} ? 1 : 0;
4837 "auto-converge" => 1,
4839 "x-rdma-pin-all" => 0,
4842 "dirty-bitmaps" => $dirty_bitmaps,
4845 my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4847 for my $supported_capability (@$supported_capabilities) {
4849 capability => $supported_capability->{capability},
4850 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4854 mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4858 my ($conf, $func, @param) = @_;
4862 my $test_volid = sub {
4863 my ($key, $drive, $snapname) = @_;
4865 my $volid = $drive->{file};
4868 $volhash->{$volid}->{cdrom} //= 1;
4869 $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4871 my $replicate = $drive->{replicate} // 1;
4872 $volhash->{$volid}->{replicate} //= 0;
4873 $volhash->{$volid}->{replicate} = 1 if $replicate;
4875 $volhash->{$volid}->{shared} //= 0;
4876 $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4878 $volhash->{$volid}->{referenced_in_config} //= 0;
4879 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
4881 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4882 if defined($snapname);
4884 my $size = $drive->{size};
4885 $volhash->{$volid}->{size} //= $size if $size;
4887 $volhash->{$volid}->{is_vmstate} //= 0;
4888 $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate
';
4890 $volhash->{$volid}->{is_tpmstate} //= 0;
4891 $volhash->{$volid}->{is_tpmstate} = 1 if $key eq 'tpmstate0
';
4893 $volhash->{$volid}->{is_unused} //= 0;
4894 $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4896 $volhash->{$volid}->{drivename} = $key if is_valid_drivename($key);
4899 my $include_opts = {
4900 extra_keys => ['vmstate
'],
4901 include_unused => 1,
4904 PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4905 foreach my $snapname (keys %{$conf->{snapshots}}) {
4906 my $snap = $conf->{snapshots}->{$snapname};
4907 PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4910 foreach my $volid (keys %$volhash) {
4911 &$func($volid, $volhash->{$volid}, @param);
4915 my $fast_plug_option = {
4919 'migrate_downtime
' => 1,
4920 'migrate_speed
' => 1,
4927 'vmstatestorage
' => 1,
4930 for my $opt (keys %$confdesc_cloudinit) {
4931 $fast_plug_option->{$opt} = 1;
4934 # hotplug changes in [PENDING]
4935 # $selection hash can be used to only apply specified options, for
4936 # example: { cores => 1 } (only apply changed 'cores
')
4937 # $errors ref is used to return error messages
4938 sub vmconfig_hotplug_pending {
4939 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4941 my $defaults = load_defaults();
4942 my $arch = get_vm_arch($conf);
4943 my $machine_type = get_vm_machine($conf, undef, $arch);
4945 # commit values which do not have any impact on running VM first
4946 # Note: those option cannot raise errors, we we do not care about
4947 # $selection and always apply them.
4949 my $add_error = sub {
4950 my ($opt, $msg) = @_;
4951 $errors->{$opt} = "hotplug problem - $msg";
4954 my $cloudinit_pending_properties = PVE::QemuServer::cloudinit_pending_properties();
4956 my $cloudinit_record_changed = sub {
4957 my ($conf, $opt, $old, $new) = @_;
4958 return if !$cloudinit_pending_properties->{$opt};
4960 my $ci = ($conf->{cloudinit} //= {});
4962 my $recorded = $ci->{$opt};
4963 my %added = map { $_ => 1 } PVE::Tools::split_list(delete($ci->{added}) // '');
4965 if (defined($new)) {
4966 if (defined($old)) {
4967 # an existing value is being modified
4968 if (defined($recorded)) {
4969 # the value was already not in sync
4970 if ($new eq $recorded) {
4971 # a value is being reverted to the cloud-init state:
4973 delete $added{$opt};
4975 # the value was changed multiple times, do nothing
4977 } elsif ($added{$opt}) {
4978 # the value had been marked as added and is being changed, do nothing
4980 # the value is new, record it:
4984 # a new value is being added
4985 if (defined($recorded)) {
4986 # it was already not in sync
4987 if ($new eq $recorded) {
4988 # a value is being reverted to the cloud-init state:
4990 delete $added{$opt};
4992 # the value had temporarily been removed, do nothing
4994 } elsif ($added{$opt}) {
4995 # the value had been marked as added already, do nothing
4997 # the value is new, add it
5001 } elsif (!defined($old)) {
5002 # a non-existent value is being removed? ignore...
5004 # a value is being deleted
5005 if (defined($recorded)) {
5006 # a value was already recorded, just keep it
5007 } elsif ($added{$opt}) {
5008 # the value was marked as added, remove it
5009 delete $added{$opt};
5011 # a previously unrecorded value is being removed, record the old value:
5016 my $added = join(',', sort keys %added);
5017 $ci->{added} = $added if length($added);
5021 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5022 if ($fast_plug_option->{$opt}) {
5023 my $new = delete $conf->{pending}->{$opt};
5024 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $new);
5025 $conf->{$opt} = $new;
5031 PVE::QemuConfig->write_config($vmid, $conf);
5034 my $ostype = $conf->{ostype};
5035 my $version = extract_version($machine_type, get_running_qemu_version($vmid));
5036 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
5037 my $usb_hotplug = $hotplug_features->{usb}
5038 && min_version($version, 7, 1)
5039 && defined($ostype) && ($ostype eq 'l26
' || windows_version($ostype) > 7);
5041 my $cgroup = PVE::QemuServer::CGroup->new($vmid);
5042 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
5044 foreach my $opt (sort keys %$pending_delete_hash) {
5045 next if $selection && !$selection->{$opt};
5046 my $force = $pending_delete_hash->{$opt}->{force};
5048 if ($opt eq 'hotplug
') {
5049 die "skip\n" if ($conf->{hotplug} =~ /memory/);
5050 } elsif ($opt eq 'tablet
') {
5051 die "skip\n" if !$hotplug_features->{usb};
5052 if ($defaults->{tablet}) {
5053 vm_deviceplug($storecfg, $conf, $vmid, 'tablet
', $arch, $machine_type);
5054 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard
', $arch, $machine_type)
5055 if $arch eq 'aarch64
';
5057 vm_deviceunplug($vmid, $conf, 'tablet
');
5058 vm_deviceunplug($vmid, $conf, 'keyboard
') if $arch eq 'aarch64
';
5060 } elsif ($opt =~ m/^usb(\d+)$/) {
5062 die "skip\n" if !$usb_hotplug;
5063 vm_deviceunplug($vmid, $conf, "usbredirdev$index"); # if it's a spice port
5064 vm_deviceunplug
($vmid, $conf, $opt);
5065 } elsif ($opt eq 'vcpus') {
5066 die "skip\n" if !$hotplug_features->{cpu
};
5067 qemu_cpu_hotplug
($vmid, $conf, undef);
5068 } elsif ($opt eq 'balloon') {
5069 # enable balloon device is not hotpluggable
5070 die "skip\n" if defined($conf->{balloon
}) && $conf->{balloon
} == 0;
5071 # here we reset the ballooning value to memory
5072 my $balloon = $conf->{memory
} || $defaults->{memory
};
5073 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5074 } elsif ($fast_plug_option->{$opt}) {
5076 } elsif ($opt =~ m/^net(\d+)$/) {
5077 die "skip\n" if !$hotplug_features->{network
};
5078 vm_deviceunplug
($vmid, $conf, $opt);
5079 } elsif (is_valid_drivename
($opt)) {
5080 die "skip\n" if !$hotplug_features->{disk
} || $opt =~ m/(ide|sata)(\d+)/;
5081 vm_deviceunplug
($vmid, $conf, $opt);
5082 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5083 } elsif ($opt =~ m/^memory$/) {
5084 die "skip\n" if !$hotplug_features->{memory
};
5085 PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults);
5086 } elsif ($opt eq 'cpuunits') {
5087 $cgroup->change_cpu_shares(undef);
5088 } elsif ($opt eq 'cpulimit') {
5089 $cgroup->change_cpu_quota(undef, undef); # reset, cgroup module can better decide values
5095 &$add_error($opt, $err) if $err ne "skip\n";
5097 my $old = delete $conf->{$opt};
5098 $cloudinit_record_changed->($conf, $opt, $old, undef);
5099 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5104 foreach my $opt (keys %{$conf->{pending
}}) {
5105 next if $selection && !$selection->{$opt};
5106 my $value = $conf->{pending
}->{$opt};
5108 if ($opt eq 'hotplug') {
5109 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug
} =~ /memory/);
5110 } elsif ($opt eq 'tablet') {
5111 die "skip\n" if !$hotplug_features->{usb
};
5113 vm_deviceplug
($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5114 vm_deviceplug
($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5115 if $arch eq 'aarch64';
5116 } elsif ($value == 0) {
5117 vm_deviceunplug
($vmid, $conf, 'tablet');
5118 vm_deviceunplug
($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5120 } elsif ($opt =~ m/^usb(\d+)$/) {
5122 die "skip\n" if !$usb_hotplug;
5123 my $d = eval { parse_property_string
($usbdesc->{format
}, $value) };
5125 if ($d->{host
} eq 'spice') {
5126 $id = "usbredirdev$index";
5128 qemu_usb_hotplug
($storecfg, $conf, $vmid, $id, $d, $arch, $machine_type);
5129 } elsif ($opt eq 'vcpus') {
5130 die "skip\n" if !$hotplug_features->{cpu
};
5131 qemu_cpu_hotplug
($vmid, $conf, $value);
5132 } elsif ($opt eq 'balloon') {
5133 # enable/disable balloning device is not hotpluggable
5134 my $old_balloon_enabled = !!(!defined($conf->{balloon
}) || $conf->{balloon
});
5135 my $new_balloon_enabled = !!(!defined($conf->{pending
}->{balloon
}) || $conf->{pending
}->{balloon
});
5136 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5138 # allow manual ballooning if shares is set to zero
5139 if ((defined($conf->{shares
}) && ($conf->{shares
} == 0))) {
5140 my $balloon = $conf->{pending
}->{balloon
} || $conf->{memory
} || $defaults->{memory
};
5141 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5143 } elsif ($opt =~ m/^net(\d+)$/) {
5144 # some changes can be done without hotplug
5145 vmconfig_update_net
($storecfg, $conf, $hotplug_features->{network
},
5146 $vmid, $opt, $value, $arch, $machine_type);
5147 } elsif (is_valid_drivename
($opt)) {
5148 die "skip\n" if $opt eq 'efidisk0' || $opt eq 'tpmstate0';
5149 # some changes can be done without hotplug
5150 my $drive = parse_drive
($opt, $value);
5151 if (drive_is_cloudinit
($drive)) {
5152 $cloudinit_opt = [$opt, $drive];
5153 # apply all the other changes first, then generate the cloudinit disk
5156 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5157 $vmid, $opt, $value, $arch, $machine_type);
5158 } elsif ($opt =~ m/^memory$/) { #dimms
5159 die "skip\n" if !$hotplug_features->{memory
};
5160 $value = PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults, $value);
5161 } elsif ($opt eq 'cpuunits') {
5162 my $new_cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{pending
}->{$opt}); #clamp
5163 $cgroup->change_cpu_shares($new_cpuunits);
5164 } elsif ($opt eq 'cpulimit') {
5165 my $cpulimit = $conf->{pending
}->{$opt} == 0 ?
-1 : int($conf->{pending
}->{$opt} * 100000);
5166 $cgroup->change_cpu_quota($cpulimit, 100000);
5167 } elsif ($opt eq 'agent') {
5168 vmconfig_update_agent
($conf, $opt, $value);
5170 die "skip\n"; # skip non-hot-pluggable options
5174 &$add_error($opt, $err) if $err ne "skip\n";
5176 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $value);
5177 $conf->{$opt} = $value;
5178 delete $conf->{pending
}->{$opt};
5182 if (defined($cloudinit_opt)) {
5183 my ($opt, $drive) = @$cloudinit_opt;
5184 my $value = $conf->{pending
}->{$opt};
5186 my $temp = {%$conf, $opt => $value};
5187 PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($temp, $vmid);
5188 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5189 $vmid, $opt, $value, $arch, $machine_type);
5192 &$add_error($opt, $err) if $err ne "skip\n";
5194 $conf->{$opt} = $value;
5195 delete $conf->{pending
}->{$opt};
5199 # unplug xhci controller if no usb device is left
5202 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
5203 next if !defined($conf->{"usb$i"});
5208 vm_deviceunplug
($vmid, $conf, 'xhci');
5212 PVE
::QemuConfig-
>write_config($vmid, $conf);
5214 if ($hotplug_features->{cloudinit
} && PVE
::QemuServer
::Cloudinit
::has_changes
($conf)) {
5215 PVE
::QemuServer
::vmconfig_update_cloudinit_drive
($storecfg, $conf, $vmid);
5219 sub try_deallocate_drive
{
5220 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5222 if (($force || $key =~ /^unused/) && !drive_is_cdrom
($drive, 1)) {
5223 my $volid = $drive->{file
};
5224 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
5225 my $sid = PVE
::Storage
::parse_volume_id
($volid);
5226 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5228 # check if the disk is really unused
5229 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5230 if PVE
::QemuServer
::Drive
::is_volume_in_use
($storecfg, $conf, $key, $volid);
5231 PVE
::Storage
::vdisk_free
($storecfg, $volid);
5234 # If vm is not owner of this disk remove from config
5242 sub vmconfig_delete_or_detach_drive
{
5243 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5245 my $drive = parse_drive
($opt, $conf->{$opt});
5247 my $rpcenv = PVE
::RPCEnvironment
::get
();
5248 my $authuser = $rpcenv->get_user();
5251 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5252 try_deallocate_drive
($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5254 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $drive);
5260 sub vmconfig_apply_pending
{
5261 my ($vmid, $conf, $storecfg, $errors, $skip_cloud_init) = @_;
5263 return if !scalar(keys %{$conf->{pending
}});
5265 my $add_apply_error = sub {
5266 my ($opt, $msg) = @_;
5267 my $err_msg = "unable to apply pending change $opt : $msg";
5268 $errors->{$opt} = $err_msg;
5274 my $pending_delete_hash = PVE
::QemuConfig-
>parse_pending_delete($conf->{pending
}->{delete});
5275 foreach my $opt (sort keys %$pending_delete_hash) {
5276 my $force = $pending_delete_hash->{$opt}->{force
};
5278 if ($opt =~ m/^unused/) {
5279 die "internal error";
5280 } elsif (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5281 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5285 $add_apply_error->($opt, $err);
5287 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5288 delete $conf->{$opt};
5292 PVE
::QemuConfig-
>cleanup_pending($conf);
5294 my $generate_cloudinit = $skip_cloud_init ?
0 : undef;
5296 foreach my $opt (keys %{$conf->{pending
}}) { # add/change
5297 next if $opt eq 'delete'; # just to be sure
5299 if (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5300 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, parse_drive
($opt, $conf->{$opt}))
5304 $add_apply_error->($opt, $err);
5307 if (is_valid_drivename
($opt)) {
5308 my $drive = parse_drive
($opt, $conf->{pending
}->{$opt});
5309 $generate_cloudinit //= 1 if drive_is_cloudinit
($drive);
5312 $conf->{$opt} = delete $conf->{pending
}->{$opt};
5316 # write all changes at once to avoid unnecessary i/o
5317 PVE
::QemuConfig-
>write_config($vmid, $conf);
5318 if ($generate_cloudinit) {
5319 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5320 # After successful generation and if there were changes to be applied, update the
5321 # config to drop the {cloudinit} entry.
5322 PVE
::QemuConfig-
>write_config($vmid, $conf);
5327 sub vmconfig_update_net
{
5328 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5330 my $newnet = parse_net
($value);
5332 if ($conf->{$opt}) {
5333 my $oldnet = parse_net
($conf->{$opt});
5335 if (safe_string_ne
($oldnet->{model
}, $newnet->{model
}) ||
5336 safe_string_ne
($oldnet->{macaddr
}, $newnet->{macaddr
}) ||
5337 safe_num_ne
($oldnet->{queues
}, $newnet->{queues
}) ||
5338 safe_num_ne
($oldnet->{mtu
}, $newnet->{mtu
}) ||
5339 !($newnet->{bridge
} && $oldnet->{bridge
})) { # bridge/nat mode change
5341 # for non online change, we try to hot-unplug
5342 die "skip\n" if !$hotplug;
5343 vm_deviceunplug
($vmid, $conf, $opt);
5346 die "internal error" if $opt !~ m/net(\d+)/;
5347 my $iface = "tap${vmid}i$1";
5349 if (safe_string_ne
($oldnet->{bridge
}, $newnet->{bridge
}) ||
5350 safe_num_ne
($oldnet->{tag
}, $newnet->{tag
}) ||
5351 safe_string_ne
($oldnet->{trunks
}, $newnet->{trunks
}) ||
5352 safe_num_ne
($oldnet->{firewall
}, $newnet->{firewall
})) {
5353 PVE
::Network
::tap_unplug
($iface);
5356 PVE
::Network
::SDN
::Zones
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5358 PVE
::Network
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5360 } elsif (safe_num_ne
($oldnet->{rate
}, $newnet->{rate
})) {
5361 # Rate can be applied on its own but any change above needs to
5362 # include the rate in tap_plug since OVS resets everything.
5363 PVE
::Network
::tap_rate_limit
($iface, $newnet->{rate
});
5366 if (safe_string_ne
($oldnet->{link_down
}, $newnet->{link_down
})) {
5367 qemu_set_link_status
($vmid, $opt, !$newnet->{link_down
});
5375 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5381 sub vmconfig_update_agent
{
5382 my ($conf, $opt, $value) = @_;
5384 die "skip\n" if !$conf->{$opt};
5386 my $hotplug_options = { fstrim_cloned_disks
=> 1 };
5388 my $old_agent = parse_guest_agent
($conf);
5389 my $agent = parse_guest_agent
({$opt => $value});
5391 for my $option (keys %$agent) { # added/changed options
5392 next if defined($hotplug_options->{$option});
5393 die "skip\n" if safe_string_ne
($agent->{$option}, $old_agent->{$option});
5396 for my $option (keys %$old_agent) { # removed options
5397 next if defined($hotplug_options->{$option});
5398 die "skip\n" if safe_string_ne
($old_agent->{$option}, $agent->{$option});
5401 return; # either no actual change (e.g., format string reordered) or just hotpluggable changes
5404 sub vmconfig_update_disk
{
5405 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5407 my $drive = parse_drive
($opt, $value);
5409 if ($conf->{$opt} && (my $old_drive = parse_drive
($opt, $conf->{$opt}))) {
5410 my $media = $drive->{media
} || 'disk';
5411 my $oldmedia = $old_drive->{media
} || 'disk';
5412 die "unable to change media type\n" if $media ne $oldmedia;
5414 if (!drive_is_cdrom
($old_drive)) {
5416 if ($drive->{file
} ne $old_drive->{file
}) {
5418 die "skip\n" if !$hotplug;
5420 # unplug and register as unused
5421 vm_deviceunplug
($vmid, $conf, $opt);
5422 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive)
5425 # update existing disk
5427 # skip non hotpluggable value
5428 if (safe_string_ne
($drive->{aio
}, $old_drive->{aio
}) ||
5429 safe_string_ne
($drive->{discard
}, $old_drive->{discard
}) ||
5430 safe_string_ne
($drive->{iothread
}, $old_drive->{iothread
}) ||
5431 safe_string_ne
($drive->{queues
}, $old_drive->{queues
}) ||
5432 safe_string_ne
($drive->{cache
}, $old_drive->{cache
}) ||
5433 safe_string_ne
($drive->{ssd
}, $old_drive->{ssd
}) ||
5434 safe_string_ne
($drive->{ro
}, $old_drive->{ro
})) {
5439 if (safe_num_ne
($drive->{mbps
}, $old_drive->{mbps
}) ||
5440 safe_num_ne
($drive->{mbps_rd
}, $old_drive->{mbps_rd
}) ||
5441 safe_num_ne
($drive->{mbps_wr
}, $old_drive->{mbps_wr
}) ||
5442 safe_num_ne
($drive->{iops
}, $old_drive->{iops
}) ||
5443 safe_num_ne
($drive->{iops_rd
}, $old_drive->{iops_rd
}) ||
5444 safe_num_ne
($drive->{iops_wr
}, $old_drive->{iops_wr
}) ||
5445 safe_num_ne
($drive->{mbps_max
}, $old_drive->{mbps_max
}) ||
5446 safe_num_ne
($drive->{mbps_rd_max
}, $old_drive->{mbps_rd_max
}) ||
5447 safe_num_ne
($drive->{mbps_wr_max
}, $old_drive->{mbps_wr_max
}) ||
5448 safe_num_ne
($drive->{iops_max
}, $old_drive->{iops_max
}) ||
5449 safe_num_ne
($drive->{iops_rd_max
}, $old_drive->{iops_rd_max
}) ||
5450 safe_num_ne
($drive->{iops_wr_max
}, $old_drive->{iops_wr_max
}) ||
5451 safe_num_ne
($drive->{bps_max_length
}, $old_drive->{bps_max_length
}) ||
5452 safe_num_ne
($drive->{bps_rd_max_length
}, $old_drive->{bps_rd_max_length
}) ||
5453 safe_num_ne
($drive->{bps_wr_max_length
}, $old_drive->{bps_wr_max_length
}) ||
5454 safe_num_ne
($drive->{iops_max_length
}, $old_drive->{iops_max_length
}) ||
5455 safe_num_ne
($drive->{iops_rd_max_length
}, $old_drive->{iops_rd_max_length
}) ||
5456 safe_num_ne
($drive->{iops_wr_max_length
}, $old_drive->{iops_wr_max_length
})) {
5458 qemu_block_set_io_throttle
(
5460 ($drive->{mbps
} || 0)*1024*1024,
5461 ($drive->{mbps_rd
} || 0)*1024*1024,
5462 ($drive->{mbps_wr
} || 0)*1024*1024,
5463 $drive->{iops
} || 0,
5464 $drive->{iops_rd
} || 0,
5465 $drive->{iops_wr
} || 0,
5466 ($drive->{mbps_max
} || 0)*1024*1024,
5467 ($drive->{mbps_rd_max
} || 0)*1024*1024,
5468 ($drive->{mbps_wr_max
} || 0)*1024*1024,
5469 $drive->{iops_max
} || 0,
5470 $drive->{iops_rd_max
} || 0,
5471 $drive->{iops_wr_max
} || 0,
5472 $drive->{bps_max_length
} || 1,
5473 $drive->{bps_rd_max_length
} || 1,
5474 $drive->{bps_wr_max_length
} || 1,
5475 $drive->{iops_max_length
} || 1,
5476 $drive->{iops_rd_max_length
} || 1,
5477 $drive->{iops_wr_max_length
} || 1,
5487 if ($drive->{file
} eq 'none') {
5488 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5489 if (drive_is_cloudinit
($old_drive)) {
5490 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive);
5493 my $path = get_iso_path
($storecfg, $vmid, $drive->{file
});
5495 # force eject if locked
5496 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5499 mon_cmd
($vmid, "blockdev-change-medium",
5500 id
=> "$opt", filename
=> "$path");
5508 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5510 PVE
::Storage
::activate_volumes
($storecfg, [$drive->{file
}]) if $drive->{file
} !~ m
|^/dev/.+|;
5511 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5514 sub vmconfig_update_cloudinit_drive
{
5515 my ($storecfg, $conf, $vmid) = @_;
5517 my $cloudinit_ds = undef;
5518 my $cloudinit_drive = undef;
5520 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5521 my ($ds, $drive) = @_;
5522 if (PVE
::QemuServer
::drive_is_cloudinit
($drive)) {
5523 $cloudinit_ds = $ds;
5524 $cloudinit_drive = $drive;
5528 return if !$cloudinit_drive;
5530 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5531 PVE
::QemuConfig-
>write_config($vmid, $conf);
5534 my $running = PVE
::QemuServer
::check_running
($vmid);
5537 my $path = PVE
::Storage
::path
($storecfg, $cloudinit_drive->{file
});
5539 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$cloudinit_ds");
5540 mon_cmd
($vmid, "blockdev-change-medium", id
=> "$cloudinit_ds", filename
=> "$path");
5545 # called in locked context by incoming migration
5546 sub vm_migrate_get_nbd_disks
{
5547 my ($storecfg, $conf, $replicated_volumes) = @_;
5549 my $local_volumes = {};
5550 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5551 my ($ds, $drive) = @_;
5553 return if drive_is_cdrom
($drive);
5554 return if $ds eq 'tpmstate0';
5556 my $volid = $drive->{file
};
5560 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
5562 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5563 return if $scfg->{shared
};
5565 # replicated disks re-use existing state via bitmap
5566 my $use_existing = $replicated_volumes->{$volid} ?
1 : 0;
5567 $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
5569 return $local_volumes;
5572 # called in locked context by incoming migration
5573 sub vm_migrate_alloc_nbd_disks
{
5574 my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
5577 foreach my $opt (sort keys %$source_volumes) {
5578 my ($volid, $storeid, $volname, $drive, $use_existing, $format) = @{$source_volumes->{$opt}};
5580 if ($use_existing) {
5581 $nbd->{$opt}->{drivestr
} = print_drive
($drive);
5582 $nbd->{$opt}->{volid
} = $volid;
5583 $nbd->{$opt}->{replicated
} = 1;
5587 # storage mapping + volname = regular migration
5588 # storage mapping + format = remote migration
5589 # order of precedence, filtered by whether storage supports it:
5590 # 1. explicit requested format
5591 # 2. format of current volume
5592 # 3. default format of storage
5593 if (!$storagemap->{identity
}) {
5594 $storeid = PVE
::JSONSchema
::map_id
($storagemap, $storeid);
5595 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
5596 if (!$format || !grep { $format eq $_ } @$validFormats) {
5598 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5599 my $fileFormat = qemu_img_format
($scfg, $volname);
5600 $format = $fileFormat
5601 if grep { $fileFormat eq $_ } @$validFormats;
5603 $format //= $defFormat;
5606 # can't happen for remote migration, so $volname is always defined
5607 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5608 $format = qemu_img_format
($scfg, $volname);
5611 my $size = $drive->{size
} / 1024;
5612 my $newvolid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $format, undef, $size);
5613 my $newdrive = $drive;
5614 $newdrive->{format
} = $format;
5615 $newdrive->{file
} = $newvolid;
5616 my $drivestr = print_drive
($newdrive);
5617 $nbd->{$opt}->{drivestr
} = $drivestr;
5618 $nbd->{$opt}->{volid
} = $newvolid;
5624 # see vm_start_nolock for parameters, additionally:
5626 # storagemap = parsed storage map for allocating NBD disks
5628 my ($storecfg, $vmid, $params, $migrate_opts) = @_;
5630 return PVE
::QemuConfig-
>lock_config($vmid, sub {
5631 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migrate_opts->{migratedfrom
});
5633 die "you can't start a vm if it's a template\n"
5634 if !$params->{skiptemplate
} && PVE
::QemuConfig-
>is_template($conf);
5636 my $has_suspended_lock = PVE
::QemuConfig-
>has_lock($conf, 'suspended');
5637 my $has_backup_lock = PVE
::QemuConfig-
>has_lock($conf, 'backup');
5639 my $running = check_running
($vmid, undef, $migrate_opts->{migratedfrom
});
5641 if ($has_backup_lock && $running) {
5642 # a backup is currently running, attempt to start the guest in the
5643 # existing QEMU instance
5644 return vm_resume
($vmid);
5647 PVE
::QemuConfig-
>check_lock($conf)
5648 if !($params->{skiplock
} || $has_suspended_lock);
5650 $params->{resume
} = $has_suspended_lock || defined($conf->{vmstate
});
5652 die "VM $vmid already running\n" if $running;
5654 if (my $storagemap = $migrate_opts->{storagemap
}) {
5655 my $replicated = $migrate_opts->{replicated_volumes
};
5656 my $disks = vm_migrate_get_nbd_disks
($storecfg, $conf, $replicated);
5657 $migrate_opts->{nbd
} = vm_migrate_alloc_nbd_disks
($storecfg, $vmid, $disks, $storagemap);
5659 foreach my $opt (keys %{$migrate_opts->{nbd
}}) {
5660 $conf->{$opt} = $migrate_opts->{nbd
}->{$opt}->{drivestr
};
5664 return vm_start_nolock
($storecfg, $vmid, $conf, $params, $migrate_opts);
5670 # statefile => 'tcp', 'unix' for migration or path/volid for RAM state
5671 # skiplock => 0/1, skip checking for config lock
5672 # skiptemplate => 0/1, skip checking whether VM is template
5673 # forcemachine => to force QEMU machine (rollback/migration)
5674 # forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
5675 # timeout => in seconds
5676 # paused => start VM in paused state (backup)
5677 # resume => resume from hibernation
5688 # nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
5689 # migratedfrom => source node
5690 # spice_ticket => used for spice migration, passed via tunnel/stdin
5691 # network => CIDR of migration network
5692 # type => secure/insecure - tunnel over encrypted connection or plain-text
5693 # nbd_proto_version => int, 0 for TCP, 1 for UNIX
5694 # replicated_volumes => which volids should be re-used with bitmaps for nbd migration
5695 # offline_volumes => new volids of offline migrated disks like tpmstate and cloudinit, not yet
5696 # contained in config
5697 sub vm_start_nolock
{
5698 my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
5700 my $statefile = $params->{statefile
};
5701 my $resume = $params->{resume
};
5703 my $migratedfrom = $migrate_opts->{migratedfrom
};
5704 my $migration_type = $migrate_opts->{type
};
5708 # clean up leftover reboot request files
5709 eval { clear_reboot_request
($vmid); };
5712 if (!$statefile && scalar(keys %{$conf->{pending
}})) {
5713 vmconfig_apply_pending
($vmid, $conf, $storecfg);
5714 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5717 # don't regenerate the ISO if the VM is started as part of a live migration
5718 # this way we can reuse the old ISO with the correct config
5719 if (!$migratedfrom) {
5720 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5721 # FIXME: apply_cloudinit_config updates $conf in this case, and it would only drop
5722 # $conf->{cloudinit}, so we could just not do this?
5723 # But we do it above, so for now let's be consistent.
5724 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5728 # override offline migrated volumes, conf is out of date still
5729 if (my $offline_volumes = $migrate_opts->{offline_volumes
}) {
5730 for my $key (sort keys $offline_volumes->%*) {
5731 my $parsed = parse_drive
($key, $conf->{$key});
5732 $parsed->{file
} = $offline_volumes->{$key};
5733 $conf->{$key} = print_drive
($parsed);
5737 my $defaults = load_defaults
();
5739 # set environment variable useful inside network script
5740 # for remote migration the config is available on the target node!
5741 if (!$migrate_opts->{remote_node
}) {
5742 $ENV{PVE_MIGRATED_FROM
} = $migratedfrom;
5745 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-start', 1);
5747 my $forcemachine = $params->{forcemachine
};
5748 my $forcecpu = $params->{forcecpu
};
5750 # enforce machine and CPU type on suspended vm to ensure HW compatibility
5751 $forcemachine = $conf->{runningmachine
};
5752 $forcecpu = $conf->{runningcpu
};
5753 print "Resuming suspended VM\n";
5756 my ($cmd, $vollist, $spice_port) = config_to_command
($storecfg, $vmid,
5757 $conf, $defaults, $forcemachine, $forcecpu, $params->{'pbs-backing'});
5760 my $get_migration_ip = sub {
5761 my ($nodename) = @_;
5763 return $migration_ip if defined($migration_ip);
5765 my $cidr = $migrate_opts->{network
};
5767 if (!defined($cidr)) {
5768 my $dc_conf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5769 $cidr = $dc_conf->{migration
}->{network
};
5772 if (defined($cidr)) {
5773 my $ips = PVE
::Network
::get_local_ip_from_cidr
($cidr);
5775 die "could not get IP: no address configured on local " .
5776 "node for network '$cidr'\n" if scalar(@$ips) == 0;
5778 die "could not get IP: multiple addresses configured on local " .
5779 "node for network '$cidr'\n" if scalar(@$ips) > 1;
5781 $migration_ip = @$ips[0];
5784 $migration_ip = PVE
::Cluster
::remote_node_ip
($nodename, 1)
5785 if !defined($migration_ip);
5787 return $migration_ip;
5791 if ($statefile eq 'tcp') {
5792 my $migrate = $res->{migrate
} = { proto
=> 'tcp' };
5793 $migrate->{addr
} = "localhost";
5794 my $datacenterconf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5795 my $nodename = nodename
();
5797 if (!defined($migration_type)) {
5798 if (defined($datacenterconf->{migration
}->{type
})) {
5799 $migration_type = $datacenterconf->{migration
}->{type
};
5801 $migration_type = 'secure';
5805 if ($migration_type eq 'insecure') {
5806 $migrate->{addr
} = $get_migration_ip->($nodename);
5807 $migrate->{addr
} = "[$migrate->{addr}]" if Net
::IP
::ip_is_ipv6
($migrate->{addr
});
5810 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5811 $migrate->{port
} = PVE
::Tools
::next_migrate_port
($pfamily);
5812 $migrate->{uri
} = "tcp:$migrate->{addr}:$migrate->{port}";
5813 push @$cmd, '-incoming', $migrate->{uri
};
5816 } elsif ($statefile eq 'unix') {
5817 # should be default for secure migrations as a ssh TCP forward
5818 # tunnel is not deterministic reliable ready and fails regurarly
5819 # to set up in time, so use UNIX socket forwards
5820 my $migrate = $res->{migrate
} = { proto
=> 'unix' };
5821 $migrate->{addr
} = "/run/qemu-server/$vmid.migrate";
5822 unlink $migrate->{addr
};
5824 $migrate->{uri
} = "unix:$migrate->{addr}";
5825 push @$cmd, '-incoming', $migrate->{uri
};
5828 } elsif (-e
$statefile) {
5829 push @$cmd, '-loadstate', $statefile;
5831 my $statepath = PVE
::Storage
::path
($storecfg, $statefile);
5832 push @$vollist, $statefile;
5833 push @$cmd, '-loadstate', $statepath;
5835 } elsif ($params->{paused
}) {
5839 my $start_timeout = $params->{timeout
} // config_aware_timeout
($conf, $resume);
5841 my $pci_devices = {}; # host pci devices
5842 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
5843 my $dev = $conf->{"hostpci$i"} or next;
5844 $pci_devices->{$i} = parse_hostpci
($dev);
5847 # do not reserve pciid for mediated devices, sysfs will error out for duplicate assignment
5848 my $real_pci_devices = [ grep { !(defined($_->{mdev
}) && scalar($_->{pciid
}->@*) == 1) } values $pci_devices->%* ];
5850 # map to a flat list of pci ids
5851 my $pci_id_list = [ map { $_->{id
} } map { $_->{pciid
}->@* } $real_pci_devices->@* ];
5853 # reserve all PCI IDs before actually doing anything with them
5854 PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, $start_timeout);
5858 for my $id (sort keys %$pci_devices) {
5859 my $d = $pci_devices->{$id};
5860 for my $dev ($d->{pciid
}->@*) {
5861 my $info = PVE
::QemuServer
::PCI
::prepare_pci_device
($vmid, $dev->{id
}, $id, $d->{mdev
});
5863 # nvidia grid needs the qemu parameter '-uuid' set
5864 # use smbios uuid or mdev uuid as fallback for that
5865 if ($d->{mdev
} && !defined($uuid) && $info->{vendor
} eq '10de') {
5866 if (defined($conf->{smbios1
})) {
5867 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
5868 $uuid = $smbios_conf->{uuid
} if defined($smbios_conf->{uuid
});
5870 $uuid = PVE
::QemuServer
::PCI
::generate_mdev_uuid
($vmid, $id) if !defined($uuid);
5874 push @$cmd, '-uuid', $uuid if defined($uuid);
5877 eval { cleanup_pci_devices
($vmid, $conf) };
5882 PVE
::Storage
::activate_volumes
($storecfg, $vollist);
5885 run_command
(['/bin/systemctl', 'stop', "$vmid.scope"], outfunc
=> sub{}, errfunc
=> sub{});
5887 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5888 # timeout should be more than enough here...
5889 PVE
::Systemd
::wait_for_unit_removed
("$vmid.scope", 20);
5891 my $cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{cpuunits
});
5894 timeout
=> $statefile ?
undef : $start_timeout,
5899 # when migrating, prefix QEMU output so other side can pick up any
5900 # errors that might occur and show the user
5901 if ($migratedfrom) {
5902 $run_params{quiet
} = 1;
5903 $run_params{logfunc
} = sub { print "QEMU: $_[0]\n" };
5906 my %systemd_properties = (
5907 Slice
=> 'qemu.slice',
5908 KillMode
=> 'process',
5910 TimeoutStopUSec
=> ULONG_MAX
, # infinity
5913 if (PVE
::CGroup
::cgroup_mode
() == 2) {
5914 $systemd_properties{CPUWeight
} = $cpuunits;
5916 $systemd_properties{CPUShares
} = $cpuunits;
5919 if (my $cpulimit = $conf->{cpulimit
}) {
5920 $systemd_properties{CPUQuota
} = int($cpulimit * 100);
5922 $systemd_properties{timeout
} = 10 if $statefile; # setting up the scope shoul be quick
5924 my $run_qemu = sub {
5925 PVE
::Tools
::run_fork
sub {
5926 PVE
::Systemd
::enter_systemd_scope
($vmid, "Proxmox VE VM $vmid", %systemd_properties);
5929 if (my $tpm = $conf->{tpmstate0
}) {
5930 # start the TPM emulator so QEMU can connect on start
5931 $tpmpid = start_swtpm
($storecfg, $vmid, $tpm, $migratedfrom);
5934 my $exitcode = run_command
($cmd, %run_params);
5937 warn "stopping swtpm instance (pid $tpmpid) due to QEMU startup error\n";
5938 kill 'TERM', $tpmpid;
5940 die "QEMU exited with code $exitcode\n";
5945 if ($conf->{hugepages
}) {
5948 my $hotplug_features =
5949 parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
5950 my $hugepages_topology =
5951 PVE
::QemuServer
::Memory
::hugepages_topology
($conf, $hotplug_features->{memory
});
5953 my $hugepages_host_topology = PVE
::QemuServer
::Memory
::hugepages_host_topology
();
5955 PVE
::QemuServer
::Memory
::hugepages_mount
();
5956 PVE
::QemuServer
::Memory
::hugepages_allocate
($hugepages_topology, $hugepages_host_topology);
5958 eval { $run_qemu->() };
5960 PVE
::QemuServer
::Memory
::hugepages_reset
($hugepages_host_topology)
5961 if !$conf->{keephugepages
};
5965 PVE
::QemuServer
::Memory
::hugepages_pre_deallocate
($hugepages_topology)
5966 if !$conf->{keephugepages
};
5968 eval { PVE
::QemuServer
::Memory
::hugepages_update_locked
($code); };
5971 eval { $run_qemu->() };
5975 # deactivate volumes if start fails
5976 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
5978 eval { cleanup_pci_devices
($vmid, $conf) };
5981 die "start failed: $err";
5984 # re-reserve all PCI IDs now that we can know the actual VM PID
5985 my $pid = PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
5986 eval { PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, undef, $pid) };
5989 if (defined($res->{migrate
})) {
5990 print "migration listens on $res->{migrate}->{uri}\n";
5991 } elsif ($statefile) {
5992 eval { mon_cmd
($vmid, "cont"); };
5996 #start nbd server for storage migration
5997 if (my $nbd = $migrate_opts->{nbd
}) {
5998 my $nbd_protocol_version = $migrate_opts->{nbd_proto_version
} // 0;
6000 my $migrate_storage_uri;
6001 # nbd_protocol_version > 0 for unix socket support
6002 if ($nbd_protocol_version > 0 && ($migration_type eq 'secure' || $migration_type eq 'websocket')) {
6003 my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
6004 mon_cmd
($vmid, "nbd-server-start", addr
=> { type
=> 'unix', data
=> { path
=> $socket_path } } );
6005 $migrate_storage_uri = "nbd:unix:$socket_path";
6006 $res->{migrate
}->{unix_sockets
} = [$socket_path];
6008 my $nodename = nodename
();
6009 my $localip = $get_migration_ip->($nodename);
6010 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
6011 my $storage_migrate_port = PVE
::Tools
::next_migrate_port
($pfamily);
6013 mon_cmd
($vmid, "nbd-server-start", addr
=> {
6016 host
=> "${localip}",
6017 port
=> "${storage_migrate_port}",
6020 $localip = "[$localip]" if Net
::IP
::ip_is_ipv6
($localip);
6021 $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
6024 my $block_info = mon_cmd
($vmid, "query-block");
6025 $block_info = { map { $_->{device
} => $_ } $block_info->@* };
6027 foreach my $opt (sort keys %$nbd) {
6028 my $drivestr = $nbd->{$opt}->{drivestr
};
6029 my $volid = $nbd->{$opt}->{volid
};
6031 my $block_node = $block_info->{"drive-$opt"}->{inserted
}->{'node-name'};
6037 'node-name' => $block_node,
6038 writable
=> JSON
::true
,
6040 name
=> "drive-$opt", # NBD export name
6043 my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
6044 print "storage migration listens on $nbd_uri volume:$drivestr\n";
6045 print "re-using replicated volume: $opt - $volid\n"
6046 if $nbd->{$opt}->{replicated
};
6048 $res->{drives
}->{$opt} = $nbd->{$opt};
6049 $res->{drives
}->{$opt}->{nbd_uri
} = $nbd_uri;
6053 if ($migratedfrom) {
6055 set_migration_caps
($vmid);
6060 print "spice listens on port $spice_port\n";
6061 $res->{spice_port
} = $spice_port;
6062 if ($migrate_opts->{spice_ticket
}) {
6063 mon_cmd
($vmid, "set_password", protocol
=> 'spice', password
=>
6064 $migrate_opts->{spice_ticket
});
6065 mon_cmd
($vmid, "expire_password", protocol
=> 'spice', time => "+30");
6070 mon_cmd
($vmid, "balloon", value
=> $conf->{balloon
}*1024*1024)
6071 if !$statefile && $conf->{balloon
};
6073 foreach my $opt (keys %$conf) {
6074 next if $opt !~ m/^net\d+$/;
6075 my $nicconf = parse_net
($conf->{$opt});
6076 qemu_set_link_status
($vmid, $opt, 0) if $nicconf->{link_down
};
6078 add_nets_bridge_fdb
($conf, $vmid);
6081 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
6086 path
=> "machine/peripheral/balloon0",
6087 property
=> "guest-stats-polling-interval",
6091 log_warn
("could not set polling interval for ballooning - $@") if $@;
6095 print "Resumed VM, removing state\n";
6096 if (my $vmstate = $conf->{vmstate
}) {
6097 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6098 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6100 delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
6101 PVE
::QemuConfig-
>write_config($vmid, $conf);
6104 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'post-start');
6109 sub vm_commandline
{
6110 my ($storecfg, $vmid, $snapname) = @_;
6112 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6114 my ($forcemachine, $forcecpu);
6116 my $snapshot = $conf->{snapshots
}->{$snapname};
6117 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
6119 # check for machine or CPU overrides in snapshot
6120 $forcemachine = $snapshot->{runningmachine
};
6121 $forcecpu = $snapshot->{runningcpu
};
6123 $snapshot->{digest
} = $conf->{digest
}; # keep file digest for API
6128 my $defaults = load_defaults
();
6130 my $cmd = config_to_command
($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
6132 return PVE
::Tools
::cmd2string
($cmd);
6136 my ($vmid, $skiplock) = @_;
6138 PVE
::QemuConfig-
>lock_config($vmid, sub {
6140 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6142 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6144 mon_cmd
($vmid, "system_reset");
6148 sub get_vm_volumes
{
6152 foreach_volid
($conf, sub {
6153 my ($volid, $attr) = @_;
6155 return if $volid =~ m
|^/|;
6157 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
6160 push @$vollist, $volid;
6166 sub cleanup_pci_devices
{
6167 my ($vmid, $conf) = @_;
6169 foreach my $key (keys %$conf) {
6170 next if $key !~ m/^hostpci(\d+)$/;
6171 my $hostpciindex = $1;
6172 my $uuid = PVE
::SysFSTools
::generate_mdev_uuid
($vmid, $hostpciindex);
6173 my $d = parse_hostpci
($conf->{$key});
6175 # NOTE: avoid PVE::SysFSTools::pci_cleanup_mdev_device as it requires PCI ID and we
6176 # don't want to break ABI just for this two liner
6177 my $dev_sysfs_dir = "/sys/bus/mdev/devices/$uuid";
6179 # some nvidia vgpu driver versions want to clean the mdevs up themselves, and error
6180 # out when we do it first. so wait for 10 seconds and then try it
6181 my $pciid = $d->{pciid
}->[0]->{id
};
6182 my $info = PVE
::SysFSTools
::pci_device_info
("$pciid");
6183 if ($info->{vendor
} eq '10de') {
6187 PVE
::SysFSTools
::file_write
("$dev_sysfs_dir/remove", "1") if -e
$dev_sysfs_dir;
6190 PVE
::QemuServer
::PCI
::remove_pci_reservation
($vmid);
6193 sub vm_stop_cleanup
{
6194 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
6199 my $vollist = get_vm_volumes
($conf);
6200 PVE
::Storage
::deactivate_volumes
($storecfg, $vollist);
6202 if (my $tpmdrive = $conf->{tpmstate0
}) {
6203 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
6204 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
6206 PVE
::Storage
::unmap_volume
($storecfg, $tpm->{file
});
6211 foreach my $ext (qw(mon qmp pid vnc qga)) {
6212 unlink "/var/run/qemu-server/${vmid}.$ext";
6215 if ($conf->{ivshmem
}) {
6216 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
6217 # just delete it for now, VMs which have this already open do not
6218 # are affected, but new VMs will get a separated one. If this
6219 # becomes an issue we either add some sort of ref-counting or just
6220 # add a "don't delete on stop" flag to the ivshmem format.
6221 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name
} // $vmid);
6224 cleanup_pci_devices
($vmid, $conf);
6226 vmconfig_apply_pending
($vmid, $conf, $storecfg) if $apply_pending_changes;
6228 warn $@ if $@; # avoid errors - just warn
6231 # call only in locked context
6233 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
6235 my $pid = check_running
($vmid, $nocheck);
6240 $conf = PVE
::QemuConfig-
>load_config($vmid);
6241 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6242 if (!defined($timeout) && $shutdown && $conf->{startup
}) {
6243 my $opts = PVE
::JSONSchema
::pve_parse_startup_order
($conf->{startup
});
6244 $timeout = $opts->{down
} if $opts->{down
};
6246 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-stop');
6251 if (defined($conf) && get_qga_key
($conf, 'enabled')) {
6252 mon_cmd
($vmid, "guest-shutdown", timeout
=> $timeout);
6254 mon_cmd
($vmid, "system_powerdown");
6257 mon_cmd
($vmid, "quit");
6263 $timeout = 60 if !defined($timeout);
6266 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6271 if ($count >= $timeout) {
6273 warn "VM still running - terminating now with SIGTERM\n";
6276 die "VM quit/powerdown failed - got timeout\n";
6279 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6283 if (!check_running
($vmid, $nocheck)) {
6284 warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
6288 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
6291 die "VM quit/powerdown failed\n";
6299 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6304 if ($count >= $timeout) {
6305 warn "VM still running - terminating now with SIGKILL\n";
6310 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6313 # Note: use $nocheck to skip tests if VM configuration file exists.
6314 # We need that when migration VMs to other nodes (files already moved)
6315 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
6317 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
6319 $force = 1 if !defined($force) && !$shutdown;
6322 my $pid = check_running
($vmid, $nocheck, $migratedfrom);
6323 kill 15, $pid if $pid;
6324 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migratedfrom);
6325 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 0);
6329 PVE
::QemuConfig-
>lock_config($vmid, sub {
6330 _do_vm_stop
($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
6335 my ($vmid, $timeout) = @_;
6337 PVE
::QemuConfig-
>lock_config($vmid, sub {
6340 # only reboot if running, as qmeventd starts it again on a stop event
6341 return if !check_running
($vmid);
6343 create_reboot_request
($vmid);
6345 my $storecfg = PVE
::Storage
::config
();
6346 _do_vm_stop
($storecfg, $vmid, undef, undef, $timeout, 1);
6350 # avoid that the next normal shutdown will be confused for a reboot
6351 clear_reboot_request
($vmid);
6357 # note: if using the statestorage parameter, the caller has to check privileges
6359 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
6366 PVE
::QemuConfig-
>lock_config($vmid, sub {
6368 $conf = PVE
::QemuConfig-
>load_config($vmid);
6370 my $is_backing_up = PVE
::QemuConfig-
>has_lock($conf, 'backup');
6371 PVE
::QemuConfig-
>check_lock($conf)
6372 if !($skiplock || $is_backing_up);
6374 die "cannot suspend to disk during backup\n"
6375 if $is_backing_up && $includestate;
6377 if ($includestate) {
6378 $conf->{lock} = 'suspending';
6379 my $date = strftime
("%Y-%m-%d", localtime(time()));
6380 $storecfg = PVE
::Storage
::config
();
6381 if (!$statestorage) {
6382 $statestorage = find_vmstate_storage
($conf, $storecfg);
6383 # check permissions for the storage
6384 my $rpcenv = PVE
::RPCEnvironment
::get
();
6385 if ($rpcenv->{type
} ne 'cli') {
6386 my $authuser = $rpcenv->get_user();
6387 $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
6392 $vmstate = PVE
::QemuConfig-
>__snapshot_save_vmstate(
6393 $vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
6394 $path = PVE
::Storage
::path
($storecfg, $vmstate);
6395 PVE
::QemuConfig-
>write_config($vmid, $conf);
6397 mon_cmd
($vmid, "stop");
6401 if ($includestate) {
6403 PVE
::Storage
::activate_volumes
($storecfg, [$vmstate]);
6406 set_migration_caps
($vmid, 1);
6407 mon_cmd
($vmid, "savevm-start", statefile
=> $path);
6409 my $state = mon_cmd
($vmid, "query-savevm");
6410 if (!$state->{status
}) {
6411 die "savevm not active\n";
6412 } elsif ($state->{status
} eq 'active') {
6415 } elsif ($state->{status
} eq 'completed') {
6416 print "State saved, quitting\n";
6418 } elsif ($state->{status
} eq 'failed' && $state->{error
}) {
6419 die "query-savevm failed with error '$state->{error}'\n"
6421 die "query-savevm returned status '$state->{status}'\n";
6427 PVE
::QemuConfig-
>lock_config($vmid, sub {
6428 $conf = PVE
::QemuConfig-
>load_config($vmid);
6430 # cleanup, but leave suspending lock, to indicate something went wrong
6432 mon_cmd
($vmid, "savevm-end");
6433 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6434 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6435 delete $conf->@{qw(vmstate runningmachine runningcpu)};
6436 PVE
::QemuConfig-
>write_config($vmid, $conf);
6442 die "lock changed unexpectedly\n"
6443 if !PVE
::QemuConfig-
>has_lock($conf, 'suspending');
6445 mon_cmd
($vmid, "quit");
6446 $conf->{lock} = 'suspended';
6447 PVE
::QemuConfig-
>write_config($vmid, $conf);
6452 # $nocheck is set when called as part of a migration - in this context the
6453 # location of the config file (source or target node) is not deterministic,
6454 # since migration cannot wait for pmxcfs to process the rename
6456 my ($vmid, $skiplock, $nocheck) = @_;
6458 PVE
::QemuConfig-
>lock_config($vmid, sub {
6459 my $res = mon_cmd
($vmid, 'query-status');
6460 my $resume_cmd = 'cont';
6464 $conf = eval { PVE
::QemuConfig-
>load_config($vmid) }; # try on target node
6466 my $vmlist = PVE
::Cluster
::get_vmlist
();
6467 if (exists($vmlist->{ids
}->{$vmid})) {
6468 my $node = $vmlist->{ids
}->{$vmid}->{node
};
6469 $conf = eval { PVE
::QemuConfig-
>load_config($vmid, $node) }; # try on source node
6472 PVE
::Cluster
::cfs_update
(); # vmlist was wrong, invalidate cache
6473 $conf = PVE
::QemuConfig-
>load_config($vmid); # last try on target node again
6477 $conf = PVE
::QemuConfig-
>load_config($vmid);
6480 if ($res->{status
}) {
6481 return if $res->{status
} eq 'running'; # job done, go home
6482 $resume_cmd = 'system_wakeup' if $res->{status
} eq 'suspended';
6483 $reset = 1 if $res->{status
} eq 'shutdown';
6487 PVE
::QemuConfig-
>check_lock($conf)
6488 if !($skiplock || PVE
::QemuConfig-
>has_lock($conf, 'backup'));
6492 # required if a VM shuts down during a backup and we get a resume
6493 # request before the backup finishes for example
6494 mon_cmd
($vmid, "system_reset");
6497 add_nets_bridge_fdb
($conf, $vmid) if $resume_cmd eq 'cont';
6499 mon_cmd
($vmid, $resume_cmd);
6504 my ($vmid, $skiplock, $key) = @_;
6506 PVE
::QemuConfig-
>lock_config($vmid, sub {
6508 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6510 # there is no qmp command, so we use the human monitor command
6511 my $res = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "sendkey $key");
6512 die $res if $res ne '';
6516 # vzdump restore implementaion
6518 sub tar_archive_read_firstfile
{
6519 my $archive = shift;
6521 die "ERROR: file '$archive' does not exist\n" if ! -f
$archive;
6523 # try to detect archive type first
6524 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6525 die "unable to open file '$archive'\n";
6526 my $firstfile = <$fh>;
6530 die "ERROR: archive contaions no data\n" if !$firstfile;
6536 sub tar_restore_cleanup
{
6537 my ($storecfg, $statfile) = @_;
6539 print STDERR
"starting cleanup\n";
6541 if (my $fd = IO
::File-
>new($statfile, "r")) {
6542 while (defined(my $line = <$fd>)) {
6543 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6546 if ($volid =~ m
|^/|) {
6547 unlink $volid || die 'unlink failed\n';
6549 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6551 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6553 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6555 print STDERR
"unable to parse line in statfile - $line";
6562 sub restore_file_archive
{
6563 my ($archive, $vmid, $user, $opts) = @_;
6565 return restore_vma_archive
($archive, $vmid, $user, $opts)
6568 my $info = PVE
::Storage
::archive_info
($archive);
6569 my $format = $opts->{format
} // $info->{format
};
6570 my $comp = $info->{compression
};
6572 # try to detect archive format
6573 if ($format eq 'tar') {
6574 return restore_tar_archive
($archive, $vmid, $user, $opts);
6576 return restore_vma_archive
($archive, $vmid, $user, $opts, $comp);
6580 # hepler to remove disks that will not be used after restore
6581 my $restore_cleanup_oldconf = sub {
6582 my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
6584 my $kept_disks = {};
6586 PVE
::QemuConfig-
>foreach_volume($oldconf, sub {
6587 my ($ds, $drive) = @_;
6589 return if drive_is_cdrom
($drive, 1);
6591 my $volid = $drive->{file
};
6592 return if !$volid || $volid =~ m
|^/|;
6594 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
6595 return if !$path || !$owner || ($owner != $vmid);
6597 # Note: only delete disk we want to restore
6598 # other volumes will become unused
6599 if ($virtdev_hash->{$ds}) {
6600 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
6605 $kept_disks->{$volid} = 1;
6609 # after the restore we have no snapshots anymore
6610 for my $snapname (keys $oldconf->{snapshots
}->%*) {
6611 my $snap = $oldconf->{snapshots
}->{$snapname};
6612 if ($snap->{vmstate
}) {
6613 eval { PVE
::Storage
::vdisk_free
($storecfg, $snap->{vmstate
}); };
6619 for my $volid (keys $kept_disks->%*) {
6620 eval { PVE
::Storage
::volume_snapshot_delete
($storecfg, $volid, $snapname); };
6626 # Helper to parse vzdump backup device hints
6628 # $rpcenv: Environment, used to ckeck storage permissions
6629 # $user: User ID, to check storage permissions
6630 # $storecfg: Storage configuration
6631 # $fh: the file handle for reading the configuration
6632 # $devinfo: should contain device sizes for all backu-up'ed devices
6633 # $options: backup options (pool, default storage)
6635 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
6636 my $parse_backup_hints = sub {
6637 my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
6639 my $check_storage = sub { # assert if an image can be allocate
6640 my ($storeid, $scfg) = @_;
6641 die "Content type 'images' is not available on storage '$storeid'\n"
6642 if !$scfg->{content
}->{images
};
6643 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace'])
6644 if $user ne 'root@pam';
6647 my $virtdev_hash = {};
6648 while (defined(my $line = <$fh>)) {
6649 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6650 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6651 die "archive does not contain data for drive '$virtdev'\n"
6652 if !$devinfo->{$devname};
6654 if (defined($options->{storage
})) {
6655 $storeid = $options->{storage
} || 'local';
6656 } elsif (!$storeid) {
6659 $format = 'raw' if !$format;
6660 $devinfo->{$devname}->{devname
} = $devname;
6661 $devinfo->{$devname}->{virtdev
} = $virtdev;
6662 $devinfo->{$devname}->{format
} = $format;
6663 $devinfo->{$devname}->{storeid
} = $storeid;
6665 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6666 $check_storage->($storeid, $scfg); # permission and content type check
6668 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6669 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6671 my $drive = parse_drive
($virtdev, $2);
6673 if (drive_is_cloudinit
($drive)) {
6674 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
6675 $storeid = $options->{storage
} if defined ($options->{storage
});
6676 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6677 my $format = qemu_img_format
($scfg, $volname); # has 'raw' fallback
6679 $check_storage->($storeid, $scfg); # permission and content type check
6681 $virtdev_hash->{$virtdev} = {
6683 storeid
=> $storeid,
6684 size
=> PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
,
6691 return $virtdev_hash;
6694 # Helper to allocate and activate all volumes required for a restore
6696 # $storecfg: Storage configuration
6697 # $virtdev_hash: as returned by parse_backup_hints()
6699 # Returns: { $virtdev => $volid }
6700 my $restore_allocate_devices = sub {
6701 my ($storecfg, $virtdev_hash, $vmid) = @_;
6704 foreach my $virtdev (sort keys %$virtdev_hash) {
6705 my $d = $virtdev_hash->{$virtdev};
6706 my $alloc_size = int(($d->{size
} + 1024 - 1)/1024);
6707 my $storeid = $d->{storeid
};
6708 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6710 # test if requested format is supported
6711 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
6712 my $supported = grep { $_ eq $d->{format
} } @$validFormats;
6713 $d->{format
} = $defFormat if !$supported;
6716 if ($d->{is_cloudinit
}) {
6717 $name = "vm-$vmid-cloudinit";
6718 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6719 if ($scfg->{path
}) {
6720 $name .= ".$d->{format}";
6724 my $volid = PVE
::Storage
::vdisk_alloc
(
6725 $storecfg, $storeid, $vmid, $d->{format
}, $name, $alloc_size);
6727 print STDERR
"new volume ID is '$volid'\n";
6728 $d->{volid
} = $volid;
6730 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
6732 $map->{$virtdev} = $volid;
6738 sub restore_update_config_line
{
6739 my ($cookie, $map, $line, $unique) = @_;
6741 return '' if $line =~ m/^\#qmdump\#/;
6742 return '' if $line =~ m/^\#vzdump\#/;
6743 return '' if $line =~ m/^lock:/;
6744 return '' if $line =~ m/^unused\d+:/;
6745 return '' if $line =~ m/^parent:/;
6749 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
6750 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6751 # try to convert old 1.X settings
6752 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6753 foreach my $devconfig (PVE
::Tools
::split_list
($ethcfg)) {
6754 my ($model, $macaddr) = split(/\=/, $devconfig);
6755 $macaddr = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if !$macaddr || $unique;
6758 bridge
=> "vmbr$ind",
6759 macaddr
=> $macaddr,
6761 my $netstr = print_net
($net);
6763 $res .= "net$cookie->{netcount}: $netstr\n";
6764 $cookie->{netcount
}++;
6766 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6767 my ($id, $netstr) = ($1, $2);
6768 my $net = parse_net
($netstr);
6769 $net->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if $net->{macaddr
};
6770 $netstr = print_net
($net);
6771 $res .= "$id: $netstr\n";
6772 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk|tpmstate)\d+):\s*(\S+)\s*$/) {
6775 my $di = parse_drive
($virtdev, $value);
6776 if (defined($di->{backup
}) && !$di->{backup
}) {
6778 } elsif ($map->{$virtdev}) {
6779 delete $di->{format
}; # format can change on restore
6780 $di->{file
} = $map->{$virtdev};
6781 $value = print_drive
($di);
6782 $res .= "$virtdev: $value\n";
6786 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6788 if ($vmgenid ne '0') {
6789 # always generate a new vmgenid if there was a valid one setup
6790 $vmgenid = generate_uuid
();
6792 $res .= "vmgenid: $vmgenid\n";
6793 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6794 my ($uuid, $uuid_str);
6795 UUID
::generate
($uuid);
6796 UUID
::unparse
($uuid, $uuid_str);
6797 my $smbios1 = parse_smbios1
($2);
6798 $smbios1->{uuid
} = $uuid_str;
6799 $res .= $1.print_smbios1
($smbios1)."\n";
6807 my $restore_deactivate_volumes = sub {
6808 my ($storecfg, $virtdev_hash) = @_;
6811 for my $dev (values $virtdev_hash->%*) {
6812 push $vollist->@*, $dev->{volid
} if $dev->{volid
};
6815 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
6816 print STDERR
$@ if $@;
6819 my $restore_destroy_volumes = sub {
6820 my ($storecfg, $virtdev_hash) = @_;
6822 for my $dev (values $virtdev_hash->%*) {
6823 my $volid = $dev->{volid
} or next;
6825 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6826 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6828 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6832 my $restore_merge_config = sub {
6833 my ($filename, $backup_conf_raw, $override_conf) = @_;
6835 my $backup_conf = parse_vm_config
($filename, $backup_conf_raw);
6836 for my $key (keys $override_conf->%*) {
6837 $backup_conf->{$key} = $override_conf->{$key};
6840 return $backup_conf;
6844 my ($cfg, $vmid) = @_;
6846 my $info = PVE
::Storage
::vdisk_list
($cfg, undef, $vmid, undef, 'images');
6848 my $volid_hash = {};
6849 foreach my $storeid (keys %$info) {
6850 foreach my $item (@{$info->{$storeid}}) {
6851 next if !($item->{volid
} && $item->{size
});
6852 $item->{path
} = PVE
::Storage
::path
($cfg, $item->{volid
});
6853 $volid_hash->{$item->{volid
}} = $item;
6860 sub update_disk_config
{
6861 my ($vmid, $conf, $volid_hash) = @_;
6864 my $prefix = "VM $vmid";
6866 # used and unused disks
6867 my $referenced = {};
6869 # Note: it is allowed to define multiple storages with same path (alias), so
6870 # we need to check both 'volid' and real 'path' (two different volid can point
6871 # to the same path).
6873 my $referencedpath = {};
6876 PVE
::QemuConfig-
>foreach_volume($conf, sub {
6877 my ($opt, $drive) = @_;
6879 my $volid = $drive->{file
};
6881 my $volume = $volid_hash->{$volid};
6883 # mark volid as "in-use" for next step
6884 $referenced->{$volid} = 1;
6885 if ($volume && (my $path = $volume->{path
})) {
6886 $referencedpath->{$path} = 1;
6889 return if drive_is_cdrom
($drive);
6892 my ($updated, $msg) = PVE
::QemuServer
::Drive
::update_disksize
($drive, $volume->{size
});
6893 if (defined($updated)) {
6895 $conf->{$opt} = print_drive
($updated);
6896 print "$prefix ($opt): $msg\n";
6900 # remove 'unusedX' entry if volume is used
6901 PVE
::QemuConfig-
>foreach_unused_volume($conf, sub {
6902 my ($opt, $drive) = @_;
6904 my $volid = $drive->{file
};
6908 $path = $volid_hash->{$volid}->{path
} if $volid_hash->{$volid};
6909 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6910 print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
6912 delete $conf->{$opt};
6915 $referenced->{$volid} = 1;
6916 $referencedpath->{$path} = 1 if $path;
6919 foreach my $volid (sort keys %$volid_hash) {
6920 next if $volid =~ m/vm-$vmid-state-/;
6921 next if $referenced->{$volid};
6922 my $path = $volid_hash->{$volid}->{path
};
6923 next if !$path; # just to be sure
6924 next if $referencedpath->{$path};
6926 my $key = PVE
::QemuConfig-
>add_unused_volume($conf, $volid);
6927 print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
6928 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6935 my ($vmid, $nolock, $dryrun) = @_;
6937 my $cfg = PVE
::Storage
::config
();
6939 print "rescan volumes...\n";
6940 my $volid_hash = scan_volids
($cfg, $vmid);
6942 my $updatefn = sub {
6945 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6947 PVE
::QemuConfig-
>check_lock($conf);
6950 foreach my $volid (keys %$volid_hash) {
6951 my $info = $volid_hash->{$volid};
6952 $vm_volids->{$volid} = $info if $info->{vmid
} && $info->{vmid
} == $vmid;
6955 my $changes = update_disk_config
($vmid, $conf, $vm_volids);
6957 PVE
::QemuConfig-
>write_config($vmid, $conf) if $changes && !$dryrun;
6960 if (defined($vmid)) {
6964 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6967 my $vmlist = config_list
();
6968 foreach my $vmid (keys %$vmlist) {
6972 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6978 sub restore_proxmox_backup_archive
{
6979 my ($archive, $vmid, $user, $options) = @_;
6981 my $storecfg = PVE
::Storage
::config
();
6983 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($archive);
6984 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6986 my $fingerprint = $scfg->{fingerprint
};
6987 my $keyfile = PVE
::Storage
::PBSPlugin
::pbs_encryption_key_file_name
($storecfg, $storeid);
6989 my $repo = PVE
::PBSClient
::get_repository
($scfg);
6990 my $namespace = $scfg->{namespace
};
6992 # This is only used for `pbs-restore` and the QEMU PBS driver (live-restore)
6993 my $password = PVE
::Storage
::PBSPlugin
::pbs_get_password
($scfg, $storeid);
6994 local $ENV{PBS_PASSWORD
} = $password;
6995 local $ENV{PBS_FINGERPRINT
} = $fingerprint if defined($fingerprint);
6997 my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6998 PVE
::Storage
::parse_volname
($storecfg, $archive);
7000 die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
7002 die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
7004 my $tmpdir = "/var/tmp/vzdumptmp$$";
7008 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7009 # disable interrupts (always do cleanups)
7013 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7015 # Note: $oldconf is undef if VM does not exists
7016 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7017 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7018 my $new_conf_raw = '';
7020 my $rpcenv = PVE
::RPCEnvironment
::get
();
7021 my $devinfo = {}; # info about drives included in backup
7022 my $virtdev_hash = {}; # info about allocated drives
7030 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7032 my $cfgfn = "$tmpdir/qemu-server.conf";
7033 my $firewall_config_fn = "$tmpdir/fw.conf";
7034 my $index_fn = "$tmpdir/index.json";
7036 my $cmd = "restore";
7038 my $param = [$pbs_backup_name, "index.json", $index_fn];
7039 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7040 my $index = PVE
::Tools
::file_get_contents
($index_fn);
7041 $index = decode_json
($index);
7043 foreach my $info (@{$index->{files
}}) {
7044 if ($info->{filename
} =~ m/^(drive-\S+).img.fidx$/) {
7046 if ($info->{size
} =~ m/^(\d+)$/) { # untaint size
7047 $devinfo->{$devname}->{size
} = $1;
7049 die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
7054 my $is_qemu_server_backup = scalar(
7055 grep { $_->{filename
} eq 'qemu-server.conf.blob' } @{$index->{files
}}
7057 if (!$is_qemu_server_backup) {
7058 die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
7060 my $has_firewall_config = scalar(grep { $_->{filename
} eq 'fw.conf.blob' } @{$index->{files
}});
7062 $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
7063 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7065 if ($has_firewall_config) {
7066 $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
7067 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7069 my $pve_firewall_dir = '/etc/pve/firewall';
7070 mkdir $pve_firewall_dir; # make sure the dir exists
7071 PVE
::Tools
::file_copy
($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
7074 my $fh = IO
::File-
>new($cfgfn, "r") ||
7075 die "unable to read qemu-server.conf - $!\n";
7077 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
7079 # fixme: rate limit?
7081 # create empty/temp config
7082 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\nlock: create");
7084 $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
7087 my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
7089 foreach my $virtdev (sort keys %$virtdev_hash) {
7090 my $d = $virtdev_hash->{$virtdev};
7091 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7093 # this fails if storage is unavailable
7094 my $volid = $d->{volid
};
7095 my $path = PVE
::Storage
::path
($storecfg, $volid);
7097 # for live-restore we only want to preload the efidisk and TPM state
7098 next if $options->{live
} && $virtdev ne 'efidisk0' && $virtdev ne 'tpmstate0';
7101 if (defined(my $ns = $scfg->{namespace
})) {
7102 @ns_arg = ('--ns', $ns);
7105 my $pbs_restore_cmd = [
7106 '/usr/bin/pbs-restore',
7107 '--repository', $repo,
7110 "$d->{devname}.img.fidx",
7115 push @$pbs_restore_cmd, '--format', $d->{format
} if $d->{format
};
7116 push @$pbs_restore_cmd, '--keyfile', $keyfile if -e
$keyfile;
7118 if (PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $volid)) {
7119 push @$pbs_restore_cmd, '--skip-zero';
7122 my $dbg_cmdstring = PVE
::Tools
::cmd2string
($pbs_restore_cmd);
7123 print "restore proxmox backup image: $dbg_cmdstring\n";
7124 run_command
($pbs_restore_cmd);
7127 $fh->seek(0, 0) || die "seek failed - $!\n";
7129 my $cookie = { netcount
=> 0 };
7130 while (defined(my $line = <$fh>)) {
7131 $new_conf_raw .= restore_update_config_line
(
7143 if ($err || !$options->{live
}) {
7144 $restore_deactivate_volumes->($storecfg, $virtdev_hash);
7150 $restore_destroy_volumes->($storecfg, $virtdev_hash);
7154 if ($options->{live
}) {
7155 # keep lock during live-restore
7156 $new_conf_raw .= "\nlock: create";
7159 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $options->{override_conf
});
7160 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7162 eval { rescan
($vmid, 1); };
7165 PVE
::AccessControl
::add_vm_to_pool
($vmid, $options->{pool
}) if $options->{pool
};
7167 if ($options->{live
}) {
7173 local $SIG{PIPE
} = sub { die "got signal ($!) - abort\n"; };
7175 my $conf = PVE
::QemuConfig-
>load_config($vmid);
7176 die "cannot do live-restore for template\n" if PVE
::QemuConfig-
>is_template($conf);
7178 # these special drives are already restored before start
7179 delete $devinfo->{'drive-efidisk0'};
7180 delete $devinfo->{'drive-tpmstate0-backup'};
7184 keyfile
=> $keyfile,
7185 snapshot
=> $pbs_backup_name,
7186 namespace
=> $namespace,
7188 pbs_live_restore
($vmid, $conf, $storecfg, $devinfo, $pbs_opts);
7190 PVE
::QemuConfig-
>remove_lock($vmid, "create");
7194 sub pbs_live_restore
{
7195 my ($vmid, $conf, $storecfg, $restored_disks, $opts) = @_;
7197 print "starting VM for live-restore\n";
7198 print "repository: '$opts->{repo}', snapshot: '$opts->{snapshot}'\n";
7200 my $pbs_backing = {};
7201 for my $ds (keys %$restored_disks) {
7202 $ds =~ m/^drive-(.*)$/;
7204 $pbs_backing->{$confname} = {
7205 repository
=> $opts->{repo
},
7206 snapshot
=> $opts->{snapshot
},
7207 archive
=> "$ds.img.fidx",
7209 $pbs_backing->{$confname}->{keyfile
} = $opts->{keyfile
} if -e
$opts->{keyfile
};
7210 $pbs_backing->{$confname}->{namespace
} = $opts->{namespace
} if defined($opts->{namespace
});
7212 my $drive = parse_drive
($confname, $conf->{$confname});
7213 print "restoring '$ds' to '$drive->{file}'\n";
7216 my $drives_streamed = 0;
7218 # make sure HA doesn't interrupt our restore by stopping the VM
7219 if (PVE
::HA
::Config
::vm_is_ha_managed
($vmid)) {
7220 run_command
(['ha-manager', 'set', "vm:$vmid", '--state', 'started']);
7223 # start VM with backing chain pointing to PBS backup, environment vars for PBS driver
7224 # in QEMU (PBS_PASSWORD and PBS_FINGERPRINT) are already set by our caller
7225 vm_start_nolock
($storecfg, $vmid, $conf, {paused
=> 1, 'pbs-backing' => $pbs_backing}, {});
7227 my $qmeventd_fd = register_qmeventd_handle
($vmid);
7229 # begin streaming, i.e. data copy from PBS to target disk for every vol,
7230 # this will effectively collapse the backing image chain consisting of
7231 # [target <- alloc-track -> PBS snapshot] to just [target] (alloc-track
7232 # removes itself once all backing images vanish with 'auto-remove=on')
7234 for my $ds (sort keys %$restored_disks) {
7235 my $job_id = "restore-$ds";
7236 mon_cmd
($vmid, 'block-stream',
7237 'job-id' => $job_id,
7240 $jobs->{$job_id} = {};
7243 mon_cmd
($vmid, 'cont');
7244 qemu_drive_mirror_monitor
($vmid, undef, $jobs, 'auto', 0, 'stream');
7246 print "restore-drive jobs finished successfully, removing all tracking block devices"
7247 ." to disconnect from Proxmox Backup Server\n";
7249 for my $ds (sort keys %$restored_disks) {
7250 mon_cmd
($vmid, 'blockdev-del', 'node-name' => "$ds-pbs");
7253 close($qmeventd_fd);
7259 warn "An error occurred during live-restore: $err\n";
7260 _do_vm_stop
($storecfg, $vmid, 1, 1, 10, 0, 1);
7261 die "live-restore failed\n";
7265 sub restore_vma_archive
{
7266 my ($archive, $vmid, $user, $opts, $comp) = @_;
7268 my $readfrom = $archive;
7270 my $cfg = PVE
::Storage
::config
();
7272 my $bwlimit = $opts->{bwlimit
};
7274 my $dbg_cmdstring = '';
7275 my $add_pipe = sub {
7277 push @$commands, $cmd;
7278 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
7279 $dbg_cmdstring .= PVE
::Tools
::cmd2string
($cmd);
7284 if ($archive eq '-') {
7287 # If we use a backup from a PVE defined storage we also consider that
7288 # storage's rate limit:
7289 my (undef, $volid) = PVE
::Storage
::path_to_volume_id
($cfg, $archive);
7290 if (defined($volid)) {
7291 my ($sid, undef) = PVE
::Storage
::parse_volume_id
($volid);
7292 my $readlimit = PVE
::Storage
::get_bandwidth_limit
('restore', [$sid], $bwlimit);
7294 print STDERR
"applying read rate limit: $readlimit\n";
7295 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
7296 $add_pipe->($cstream);
7302 my $info = PVE
::Storage
::decompressor_info
('vma', $comp);
7303 my $cmd = $info->{decompressor
};
7304 push @$cmd, $readfrom;
7308 my $tmpdir = "/var/tmp/vzdumptmp$$";
7311 # disable interrupts (always do cleanups)
7315 local $SIG{HUP
} = sub { warn "got interrupt - ignored\n"; };
7317 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
7318 POSIX
::mkfifo
($mapfifo, 0600);
7320 my $openfifo = sub { open($fifofh, '>', $mapfifo) or die $! };
7322 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
7327 my $devinfo = {}; # info about drives included in backup
7328 my $virtdev_hash = {}; # info about allocated drives
7330 my $rpcenv = PVE
::RPCEnvironment
::get
();
7332 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7334 # Note: $oldconf is undef if VM does not exist
7335 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7336 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7337 my $new_conf_raw = '';
7341 my $print_devmap = sub {
7342 my $cfgfn = "$tmpdir/qemu-server.conf";
7344 # we can read the config - that is already extracted
7345 my $fh = IO
::File-
>new($cfgfn, "r") ||
7346 die "unable to read qemu-server.conf - $!\n";
7348 my $fwcfgfn = "$tmpdir/qemu-server.fw";
7350 my $pve_firewall_dir = '/etc/pve/firewall';
7351 mkdir $pve_firewall_dir; # make sure the dir exists
7352 PVE
::Tools
::file_copy
($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
7355 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
7357 foreach my $info (values %{$virtdev_hash}) {
7358 my $storeid = $info->{storeid
};
7359 next if defined($storage_limits{$storeid});
7361 my $limit = PVE
::Storage
::get_bandwidth_limit
('restore', [$storeid], $bwlimit) // 0;
7362 print STDERR
"rate limit for storage $storeid: $limit KiB/s\n" if $limit;
7363 $storage_limits{$storeid} = $limit * 1024;
7366 foreach my $devname (keys %$devinfo) {
7367 die "found no device mapping information for device '$devname'\n"
7368 if !$devinfo->{$devname}->{virtdev
};
7371 # create empty/temp config
7373 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\n");
7374 $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
7378 my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
7380 # print restore information to $fifofh
7381 foreach my $virtdev (sort keys %$virtdev_hash) {
7382 my $d = $virtdev_hash->{$virtdev};
7383 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7385 my $storeid = $d->{storeid
};
7386 my $volid = $d->{volid
};
7389 if (my $limit = $storage_limits{$storeid}) {
7390 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
7393 my $write_zeros = 1;
7394 if (PVE
::Storage
::volume_has_feature
($cfg, 'sparseinit', $volid)) {
7398 my $path = PVE
::Storage
::path
($cfg, $volid);
7400 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
7402 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
7405 $fh->seek(0, 0) || die "seek failed - $!\n";
7407 my $cookie = { netcount
=> 0 };
7408 while (defined(my $line = <$fh>)) {
7409 $new_conf_raw .= restore_update_config_line
(
7426 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7427 local $SIG{ALRM
} = sub { die "got timeout\n"; };
7429 $oldtimeout = alarm($timeout);
7436 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
7437 my ($dev_id, $size, $devname) = ($1, $2, $3);
7438 $devinfo->{$devname} = { size
=> $size, dev_id
=> $dev_id };
7439 } elsif ($line =~ m/^CTIME: /) {
7440 # we correctly received the vma config, so we can disable
7441 # the timeout now for disk allocation (set to 10 minutes, so
7442 # that we always timeout if something goes wrong)
7445 print $fifofh "done\n";
7446 my $tmp = $oldtimeout || 0;
7447 $oldtimeout = undef;
7454 print "restore vma archive: $dbg_cmdstring\n";
7455 run_command
($commands, input
=> $input, outfunc
=> $parser, afterfork
=> $openfifo);
7459 alarm($oldtimeout) if $oldtimeout;
7461 $restore_deactivate_volumes->($cfg, $virtdev_hash);
7463 close($fifofh) if $fifofh;
7468 $restore_destroy_volumes->($cfg, $virtdev_hash);
7472 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $opts->{override_conf
});
7473 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7475 eval { rescan
($vmid, 1); };
7478 PVE
::AccessControl
::add_vm_to_pool
($vmid, $opts->{pool
}) if $opts->{pool
};
7481 sub restore_tar_archive
{
7482 my ($archive, $vmid, $user, $opts) = @_;
7484 if (scalar(keys $opts->{override_conf
}->%*) > 0) {
7485 my $keystring = join(' ', keys $opts->{override_conf
}->%*);
7486 die "cannot pass along options ($keystring) when restoring from tar archive\n";
7489 if ($archive ne '-') {
7490 my $firstfile = tar_archive_read_firstfile
($archive);
7491 die "ERROR: file '$archive' does not look like a QemuServer vzdump backup\n"
7492 if $firstfile ne 'qemu-server.conf';
7495 my $storecfg = PVE
::Storage
::config
();
7497 # avoid zombie disks when restoring over an existing VM -> cleanup first
7498 # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
7499 # skiplock=1 because qmrestore has set the 'create' lock itself already
7500 my $vmcfgfn = PVE
::QemuConfig-
>config_file($vmid);
7501 destroy_vm
($storecfg, $vmid, 1, { lock => 'restore' }) if -f
$vmcfgfn;
7503 my $tocmd = "/usr/lib/qemu-server/qmextract";
7505 $tocmd .= " --storage " . PVE
::Tools
::shellquote
($opts->{storage
}) if $opts->{storage
};
7506 $tocmd .= " --pool " . PVE
::Tools
::shellquote
($opts->{pool
}) if $opts->{pool
};
7507 $tocmd .= ' --prealloc' if $opts->{prealloc
};
7508 $tocmd .= ' --info' if $opts->{info
};
7510 # tar option "xf" does not autodetect compression when read from STDIN,
7511 # so we pipe to zcat
7512 my $cmd = "zcat -f|tar xf " . PVE
::Tools
::shellquote
($archive) . " " .
7513 PVE
::Tools
::shellquote
("--to-command=$tocmd");
7515 my $tmpdir = "/var/tmp/vzdumptmp$$";
7518 local $ENV{VZDUMP_TMPDIR
} = $tmpdir;
7519 local $ENV{VZDUMP_VMID
} = $vmid;
7520 local $ENV{VZDUMP_USER
} = $user;
7522 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7523 my $new_conf_raw = '';
7525 # disable interrupts (always do cleanups)
7529 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7537 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7539 if ($archive eq '-') {
7540 print "extracting archive from STDIN\n";
7541 run_command
($cmd, input
=> "<&STDIN");
7543 print "extracting archive '$archive'\n";
7547 return if $opts->{info
};
7551 my $statfile = "$tmpdir/qmrestore.stat";
7552 if (my $fd = IO
::File-
>new($statfile, "r")) {
7553 while (defined (my $line = <$fd>)) {
7554 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
7555 $map->{$1} = $2 if $1;
7557 print STDERR
"unable to parse line in statfile - $line\n";
7563 my $confsrc = "$tmpdir/qemu-server.conf";
7565 my $srcfd = IO
::File-
>new($confsrc, "r") || die "unable to open file '$confsrc'\n";
7567 my $cookie = { netcount
=> 0 };
7568 while (defined (my $line = <$srcfd>)) {
7569 $new_conf_raw .= restore_update_config_line
(
7580 tar_restore_cleanup
($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info
};
7586 PVE
::Tools
::file_set_contents
($conffile, $new_conf_raw);
7588 PVE
::Cluster
::cfs_update
(); # make sure we read new file
7590 eval { rescan
($vmid, 1); };
7594 sub foreach_storage_used_by_vm
{
7595 my ($conf, $func) = @_;
7599 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7600 my ($ds, $drive) = @_;
7601 return if drive_is_cdrom
($drive);
7603 my $volid = $drive->{file
};
7605 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
7606 $sidhash->{$sid} = $sid if $sid;
7609 foreach my $sid (sort keys %$sidhash) {
7614 my $qemu_snap_storage = {
7617 sub do_snapshots_with_qemu
{
7618 my ($storecfg, $volid, $deviceid) = @_;
7620 return if $deviceid =~ m/tpmstate0/;
7622 my $storage_name = PVE
::Storage
::parse_volume_id
($volid);
7623 my $scfg = $storecfg->{ids
}->{$storage_name};
7624 die "could not find storage '$storage_name'\n" if !defined($scfg);
7626 if ($qemu_snap_storage->{$scfg->{type
}} && !$scfg->{krbd
}){
7630 if ($volid =~ m/\.(qcow2|qed)$/){
7637 sub qga_check_running
{
7638 my ($vmid, $nowarn) = @_;
7640 eval { mon_cmd
($vmid, "guest-ping", timeout
=> 3); };
7642 warn "QEMU Guest Agent is not running - $@" if !$nowarn;
7648 sub template_create
{
7649 my ($vmid, $conf, $disk) = @_;
7651 my $storecfg = PVE
::Storage
::config
();
7653 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7654 my ($ds, $drive) = @_;
7656 return if drive_is_cdrom
($drive);
7657 return if $disk && $ds ne $disk;
7659 my $volid = $drive->{file
};
7660 return if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
7662 my $voliddst = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
7663 $drive->{file
} = $voliddst;
7664 $conf->{$ds} = print_drive
($drive);
7665 PVE
::QemuConfig-
>write_config($vmid, $conf);
7669 sub convert_iscsi_path
{
7672 if ($path =~ m
|^iscsi
://([^/]+)/([^/]+)/(.+)$|) {
7677 my $initiator_name = get_initiator_name
();
7679 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
7680 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
7683 die "cannot convert iscsi path '$path', unkown format\n";
7686 sub qemu_img_convert
{
7687 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized, $bwlimit) = @_;
7689 my $storecfg = PVE
::Storage
::config
();
7690 my ($src_storeid, $src_volname) = PVE
::Storage
::parse_volume_id
($src_volid, 1);
7691 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid, 1);
7693 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
7697 my $src_is_iscsi = 0;
7701 PVE
::Storage
::activate_volumes
($storecfg, [$src_volid], $snapname);
7702 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7703 $src_format = qemu_img_format
($src_scfg, $src_volname);
7704 $src_path = PVE
::Storage
::path
($storecfg, $src_volid, $snapname);
7705 $src_is_iscsi = ($src_path =~ m
|^iscsi
://|);
7706 $cachemode = 'none' if $src_scfg->{type
} eq 'zfspool';
7707 } elsif (-f
$src_volid || -b
$src_volid) {
7708 $src_path = $src_volid;
7709 if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7714 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
7716 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7717 my $dst_format = qemu_img_format
($dst_scfg, $dst_volname);
7718 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7719 my $dst_is_iscsi = ($dst_path =~ m
|^iscsi
://|);
7722 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
7723 push @$cmd, '-l', "snapshot.name=$snapname"
7724 if $snapname && $src_format && $src_format eq "qcow2";
7725 push @$cmd, '-t', 'none' if $dst_scfg->{type
} eq 'zfspool';
7726 push @$cmd, '-T', $cachemode if defined($cachemode);
7727 push @$cmd, '-r', "${bwlimit}K" if defined($bwlimit);
7729 if ($src_is_iscsi) {
7730 push @$cmd, '--image-opts';
7731 $src_path = convert_iscsi_path
($src_path);
7732 } elsif ($src_format) {
7733 push @$cmd, '-f', $src_format;
7736 if ($dst_is_iscsi) {
7737 push @$cmd, '--target-image-opts';
7738 $dst_path = convert_iscsi_path
($dst_path);
7740 push @$cmd, '-O', $dst_format;
7743 push @$cmd, $src_path;
7745 if (!$dst_is_iscsi && $is_zero_initialized) {
7746 push @$cmd, "zeroinit:$dst_path";
7748 push @$cmd, $dst_path;
7753 if($line =~ m/\((\S+)\/100\
%\)/){
7755 my $transferred = int($size * $percent / 100);
7756 my $total_h = render_bytes
($size, 1);
7757 my $transferred_h = render_bytes
($transferred, 1);
7759 print "transferred $transferred_h of $total_h ($percent%)\n";
7764 eval { run_command
($cmd, timeout
=> undef, outfunc
=> $parser); };
7766 die "copy failed: $err" if $err;
7769 sub qemu_img_format
{
7770 my ($scfg, $volname) = @_;
7772 if ($scfg->{path
} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7779 sub qemu_drive_mirror
{
7780 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
7782 $jobs = {} if !$jobs;
7786 $jobs->{"drive-$drive"} = {};
7788 if ($dst_volid =~ /^nbd:/) {
7789 $qemu_target = $dst_volid;
7792 my $storecfg = PVE
::Storage
::config
();
7793 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid);
7795 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7797 $format = qemu_img_format
($dst_scfg, $dst_volname);
7799 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7801 $qemu_target = $is_zero_initialized ?
"zeroinit:$dst_path" : $dst_path;
7804 my $opts = { timeout
=> 10, device
=> "drive-$drive", mode
=> "existing", sync
=> "full", target
=> $qemu_target };
7805 $opts->{format
} = $format if $format;
7807 if (defined($src_bitmap)) {
7808 $opts->{sync
} = 'incremental';
7809 $opts->{bitmap
} = $src_bitmap;
7810 print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
7813 if (defined($bwlimit)) {
7814 $opts->{speed
} = $bwlimit * 1024;
7815 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7817 print "drive mirror is starting for drive-$drive\n";
7820 # if a job already runs for this device we get an error, catch it for cleanup
7821 eval { mon_cmd
($vmid, "drive-mirror", %$opts); };
7823 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7825 die "mirroring error: $err\n";
7828 qemu_drive_mirror_monitor
($vmid, $vmiddst, $jobs, $completion, $qga);
7831 # $completion can be either
7832 # 'complete': wait until all jobs are ready, block-job-complete them (default)
7833 # 'cancel': wait until all jobs are ready, block-job-cancel them
7834 # 'skip': wait until all jobs are ready, return with block jobs in ready state
7835 # 'auto': wait until all jobs disappear, only use for jobs which complete automatically
7836 sub qemu_drive_mirror_monitor
{
7837 my ($vmid, $vmiddst, $jobs, $completion, $qga, $op) = @_;
7839 $completion //= 'complete';
7843 my $err_complete = 0;
7845 my $starttime = time ();
7847 die "block job ('$op') timed out\n" if $err_complete > 300;
7849 my $stats = mon_cmd
($vmid, "query-block-jobs");
7852 my $running_jobs = {};
7853 for my $stat (@$stats) {
7854 next if $stat->{type
} ne $op;
7855 $running_jobs->{$stat->{device
}} = $stat;
7858 my $readycounter = 0;
7860 for my $job_id (sort keys %$jobs) {
7861 my $job = $running_jobs->{$job_id};
7863 my $vanished = !defined($job);
7864 my $complete = defined($jobs->{$job_id}->{complete
}) && $vanished;
7865 if($complete || ($vanished && $completion eq 'auto')) {
7866 print "$job_id: $op-job finished\n";
7867 delete $jobs->{$job_id};
7871 die "$job_id: '$op' has been cancelled\n" if !defined($job);
7873 my $busy = $job->{busy
};
7874 my $ready = $job->{ready
};
7875 if (my $total = $job->{len
}) {
7876 my $transferred = $job->{offset
} || 0;
7877 my $remaining = $total - $transferred;
7878 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7880 my $duration = $ctime - $starttime;
7881 my $total_h = render_bytes
($total, 1);
7882 my $transferred_h = render_bytes
($transferred, 1);
7884 my $status = sprintf(
7885 "transferred $transferred_h of $total_h ($percent%%) in %s",
7886 render_duration
($duration),
7891 $status .= ", still busy"; # shouldn't even happen? but mirror is weird
7893 $status .= ", ready";
7896 print "$job_id: $status\n" if !$jobs->{$job_id}->{ready
};
7897 $jobs->{$job_id}->{ready
} = $ready;
7900 $readycounter++ if $job->{ready
};
7903 last if scalar(keys %$jobs) == 0;
7905 if ($readycounter == scalar(keys %$jobs)) {
7906 print "all '$op' jobs are ready\n";
7908 # do the complete later (or has already been done)
7909 last if $completion eq 'skip' || $completion eq 'auto';
7911 if ($vmiddst && $vmiddst != $vmid) {
7912 my $agent_running = $qga && qga_check_running
($vmid);
7913 if ($agent_running) {
7914 print "freeze filesystem\n";
7915 eval { mon_cmd
($vmid, "guest-fsfreeze-freeze"); };
7918 print "suspend vm\n";
7919 eval { PVE
::QemuServer
::vm_suspend
($vmid, 1); };
7923 # if we clone a disk for a new target vm, we don't switch the disk
7924 PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs);
7926 if ($agent_running) {
7927 print "unfreeze filesystem\n";
7928 eval { mon_cmd
($vmid, "guest-fsfreeze-thaw"); };
7931 print "resume vm\n";
7932 eval { PVE
::QemuServer
::vm_resume
($vmid, 1, 1); };
7939 for my $job_id (sort keys %$jobs) {
7940 # try to switch the disk if source and destination are on the same guest
7941 print "$job_id: Completing block job_id...\n";
7944 if ($completion eq 'complete') {
7945 $op = 'block-job-complete';
7946 } elsif ($completion eq 'cancel') {
7947 $op = 'block-job-cancel';
7949 die "invalid completion value: $completion\n";
7951 eval { mon_cmd
($vmid, $op, device
=> $job_id) };
7952 if ($@ =~ m/cannot be completed/) {
7953 print "$job_id: block job cannot be completed, trying again.\n";
7956 print "$job_id: Completed successfully.\n";
7957 $jobs->{$job_id}->{complete
} = 1;
7968 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7969 die "block job ($op) error: $err";
7973 sub qemu_blockjobs_cancel
{
7974 my ($vmid, $jobs) = @_;
7976 foreach my $job (keys %$jobs) {
7977 print "$job: Cancelling block job\n";
7978 eval { mon_cmd
($vmid, "block-job-cancel", device
=> $job); };
7979 $jobs->{$job}->{cancel
} = 1;
7983 my $stats = mon_cmd
($vmid, "query-block-jobs");
7985 my $running_jobs = {};
7986 foreach my $stat (@$stats) {
7987 $running_jobs->{$stat->{device
}} = $stat;
7990 foreach my $job (keys %$jobs) {
7992 if (defined($jobs->{$job}->{cancel
}) && !defined($running_jobs->{$job})) {
7993 print "$job: Done.\n";
7994 delete $jobs->{$job};
7998 last if scalar(keys %$jobs) == 0;
8004 # Check for bug #4525: drive-mirror will open the target drive with the same aio setting as the
8005 # source, but some storages have problems with io_uring, sometimes even leading to crashes.
8006 my sub clone_disk_check_io_uring
{
8007 my ($src_drive, $storecfg, $src_storeid, $dst_storeid, $use_drive_mirror) = @_;
8009 return if !$use_drive_mirror;
8011 # Don't complain when not changing storage.
8012 # Assume if it works for the source, it'll work for the target too.
8013 return if $src_storeid eq $dst_storeid;
8015 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
8016 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
8018 my $cache_direct = drive_uses_cache_direct
($src_drive);
8020 my $src_uses_io_uring;
8021 if ($src_drive->{aio
}) {
8022 $src_uses_io_uring = $src_drive->{aio
} eq 'io_uring';
8024 $src_uses_io_uring = storage_allows_io_uring_default
($src_scfg, $cache_direct);
8027 die "target storage is known to cause issues with aio=io_uring (used by current drive)\n"
8028 if $src_uses_io_uring && !storage_allows_io_uring_default
($dst_scfg, $cache_direct);
8032 my ($storecfg, $source, $dest, $full, $newvollist, $jobs, $completion, $qga, $bwlimit) = @_;
8034 my ($vmid, $running) = $source->@{qw(vmid running)};
8035 my ($src_drivename, $drive, $snapname) = $source->@{qw(drivename drive snapname)};
8037 my ($newvmid, $dst_drivename, $efisize) = $dest->@{qw(vmid drivename efisize)};
8038 my ($storage, $format) = $dest->@{qw(storage format)};
8040 my $use_drive_mirror = $full && $running && $src_drivename && !$snapname;
8042 if ($src_drivename && $dst_drivename && $src_drivename ne $dst_drivename) {
8043 die "cloning from/to EFI disk requires EFI disk\n"
8044 if $src_drivename eq 'efidisk0' || $dst_drivename eq 'efidisk0';
8045 die "cloning from/to TPM state requires TPM state\n"
8046 if $src_drivename eq 'tpmstate0' || $dst_drivename eq 'tpmstate0';
8048 # This would lead to two device nodes in QEMU pointing to the same backing image!
8049 die "cannot change drive name when cloning disk from/to the same VM\n"
8050 if $use_drive_mirror && $vmid == $newvmid;
8053 die "cannot move TPM state while VM is running\n"
8054 if $use_drive_mirror && $src_drivename eq 'tpmstate0';
8058 print "create " . ($full ?
'full' : 'linked') . " clone of drive ";
8059 print "$src_drivename " if $src_drivename;
8060 print "($drive->{file})\n";
8063 $newvolid = PVE
::Storage
::vdisk_clone
($storecfg, $drive->{file
}, $newvmid, $snapname);
8064 push @$newvollist, $newvolid;
8066 my ($src_storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
8067 my $storeid = $storage || $src_storeid;
8069 my $dst_format = resolve_dst_disk_format
($storecfg, $storeid, $volname, $format);
8073 if (drive_is_cloudinit
($drive)) {
8074 $name = "vm-$newvmid-cloudinit";
8075 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8076 if ($scfg->{path
}) {
8077 $name .= ".$dst_format";
8080 $size = PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
;
8081 } elsif ($dst_drivename eq 'efidisk0') {
8082 $size = $efisize or die "internal error - need to specify EFI disk size\n";
8083 } elsif ($dst_drivename eq 'tpmstate0') {
8084 $dst_format = 'raw';
8085 $size = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8087 clone_disk_check_io_uring
($drive, $storecfg, $src_storeid, $storeid, $use_drive_mirror);
8089 $size = PVE
::Storage
::volume_size_info
($storecfg, $drive->{file
}, 10);
8091 $newvolid = PVE
::Storage
::vdisk_alloc
(
8092 $storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024)
8094 push @$newvollist, $newvolid;
8096 PVE
::Storage
::activate_volumes
($storecfg, [$newvolid]);
8098 if (drive_is_cloudinit
($drive)) {
8099 # when cloning multiple disks (e.g. during clone_vm) it might be the last disk
8100 # if this is the case, we have to complete any block-jobs still there from
8101 # previous drive-mirrors
8102 if (($completion eq 'complete') && (scalar(keys %$jobs) > 0)) {
8103 qemu_drive_mirror_monitor
($vmid, $newvmid, $jobs, $completion, $qga);
8108 my $sparseinit = PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $newvolid);
8109 if ($use_drive_mirror) {
8110 qemu_drive_mirror
($vmid, $src_drivename, $newvolid, $newvmid, $sparseinit, $jobs,
8111 $completion, $qga, $bwlimit);
8113 if ($dst_drivename eq 'efidisk0') {
8114 # the relevant data on the efidisk may be smaller than the source
8115 # e.g. on RBD/ZFS, so we use dd to copy only the amount
8116 # that is given by the OVMF_VARS.fd
8117 my $src_path = PVE
::Storage
::path
($storecfg, $drive->{file
}, $snapname);
8118 my $dst_path = PVE
::Storage
::path
($storecfg, $newvolid);
8120 my $src_format = (PVE
::Storage
::parse_volname
($storecfg, $drive->{file
}))[6];
8122 # better for Ceph if block size is not too small, see bug #3324
8125 my $cmd = ['qemu-img', 'dd', '-n', '-O', $dst_format];
8127 if ($src_format eq 'qcow2' && $snapname) {
8128 die "cannot clone qcow2 EFI disk snapshot - requires QEMU >= 6.2\n"
8129 if !min_version
(kvm_user_version
(), 6, 2);
8130 push $cmd->@*, '-l', $snapname;
8132 push $cmd->@*, "bs=$bs", "osize=$size", "if=$src_path", "of=$dst_path";
8135 qemu_img_convert
($drive->{file
}, $newvolid, $size, $snapname, $sparseinit, $bwlimit);
8141 my $size = eval { PVE
::Storage
::volume_size_info
($storecfg, $newvolid, 10) };
8143 my $disk = dclone
($drive);
8144 delete $disk->{format
};
8145 $disk->{file
} = $newvolid;
8146 $disk->{size
} = $size if defined($size);
8151 sub get_running_qemu_version
{
8153 my $res = mon_cmd
($vmid, "query-version");
8154 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
8157 sub qemu_use_old_bios_files
{
8158 my ($machine_type) = @_;
8160 return if !$machine_type;
8162 my $use_old_bios_files = undef;
8164 if ($machine_type =~ m/^(\S+)\.pxe$/) {
8166 $use_old_bios_files = 1;
8168 my $version = extract_version
($machine_type, kvm_user_version
());
8169 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
8170 # load new efi bios files on migration. So this hack is required to allow
8171 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
8172 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
8173 $use_old_bios_files = !min_version
($version, 2, 4);
8176 return ($use_old_bios_files, $machine_type);
8179 sub get_efivars_size
{
8180 my ($conf, $efidisk) = @_;
8182 my $arch = get_vm_arch
($conf);
8183 $efidisk //= $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
8184 my $smm = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
8185 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8186 return -s
$ovmf_vars;
8189 sub update_efidisk_size
{
8192 return if !defined($conf->{efidisk0
});
8194 my $disk = PVE
::QemuServer
::parse_drive
('efidisk0', $conf->{efidisk0
});
8195 $disk->{size
} = get_efivars_size
($conf);
8196 $conf->{efidisk0
} = print_drive
($disk);
8201 sub update_tpmstate_size
{
8204 my $disk = PVE
::QemuServer
::parse_drive
('tpmstate0', $conf->{tpmstate0
});
8205 $disk->{size
} = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8206 $conf->{tpmstate0
} = print_drive
($disk);
8209 sub create_efidisk
($$$$$$$) {
8210 my ($storecfg, $storeid, $vmid, $fmt, $arch, $efidisk, $smm) = @_;
8212 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8214 my $vars_size_b = -s
$ovmf_vars;
8215 my $vars_size = PVE
::Tools
::convert_size
($vars_size_b, 'b' => 'kb');
8216 my $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
8217 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
8219 qemu_img_convert
($ovmf_vars, $volid, $vars_size_b, undef, 0);
8220 my $size = PVE
::Storage
::volume_size_info
($storecfg, $volid, 3);
8222 return ($volid, $size/1024);
8225 sub vm_iothreads_list
{
8228 my $res = mon_cmd
($vmid, 'query-iothreads');
8231 foreach my $iothread (@$res) {
8232 $iothreads->{ $iothread->{id
} } = $iothread->{"thread-id"};
8239 my ($conf, $drive) = @_;
8243 if (!$conf->{scsihw
} || ($conf->{scsihw
} =~ m/^lsi/)) {
8245 } elsif ($conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single')) {
8251 my $controller = int($drive->{index} / $maxdev);
8252 my $controller_prefix = ($conf->{scsihw
} && $conf->{scsihw
} eq 'virtio-scsi-single')
8256 return ($maxdev, $controller, $controller_prefix);
8259 sub resolve_dst_disk_format
{
8260 my ($storecfg, $storeid, $src_volname, $format) = @_;
8261 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
8264 # if no target format is specified, use the source disk format as hint
8266 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8267 $format = qemu_img_format
($scfg, $src_volname);
8273 # test if requested format is supported - else use default
8274 my $supported = grep { $_ eq $format } @$validFormats;
8275 $format = $defFormat if !$supported;
8279 # NOTE: if this logic changes, please update docs & possibly gui logic
8280 sub find_vmstate_storage
{
8281 my ($conf, $storecfg) = @_;
8283 # first, return storage from conf if set
8284 return $conf->{vmstatestorage
} if $conf->{vmstatestorage
};
8286 my ($target, $shared, $local);
8288 foreach_storage_used_by_vm
($conf, sub {
8290 my $scfg = PVE
::Storage
::storage_config
($storecfg, $sid);
8291 my $dst = $scfg->{shared
} ? \
$shared : \
$local;
8292 $$dst = $sid if !$$dst || $scfg->{path
}; # prefer file based storage
8295 # second, use shared storage where VM has at least one disk
8296 # third, use local storage where VM has at least one disk
8297 # fall back to local storage
8298 $target = $shared // $local // 'local';
8304 my ($uuid, $uuid_str);
8305 UUID
::generate
($uuid);
8306 UUID
::unparse
($uuid, $uuid_str);
8310 sub generate_smbios1_uuid
{
8311 return "uuid=".generate_uuid
();
8317 mon_cmd
($vmid, 'nbd-server-stop');
8320 sub create_reboot_request
{
8322 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
8323 or die "failed to create reboot trigger file: $!\n";
8327 sub clear_reboot_request
{
8329 my $path = "/run/qemu-server/$vmid.reboot";
8332 $res = unlink($path);
8333 die "could not remove reboot request for $vmid: $!"
8334 if !$res && $! != POSIX
::ENOENT
;
8339 sub bootorder_from_legacy
{
8340 my ($conf, $bootcfg) = @_;
8342 my $boot = $bootcfg->{legacy
} || $boot_fmt->{legacy
}->{default};
8343 my $bootindex_hash = {};
8345 foreach my $o (split(//, $boot)) {
8346 $bootindex_hash->{$o} = $i*100;
8352 PVE
::QemuConfig-
>foreach_volume($conf, sub {
8353 my ($ds, $drive) = @_;
8355 if (drive_is_cdrom
($drive, 1)) {
8356 if ($bootindex_hash->{d
}) {
8357 $bootorder->{$ds} = $bootindex_hash->{d
};
8358 $bootindex_hash->{d
} += 1;
8360 } elsif ($bootindex_hash->{c
}) {
8361 $bootorder->{$ds} = $bootindex_hash->{c
}
8362 if $conf->{bootdisk
} && $conf->{bootdisk
} eq $ds;
8363 $bootindex_hash->{c
} += 1;
8367 if ($bootindex_hash->{n
}) {
8368 for (my $i = 0; $i < $MAX_NETS; $i++) {
8369 my $netname = "net$i";
8370 next if !$conf->{$netname};
8371 $bootorder->{$netname} = $bootindex_hash->{n
};
8372 $bootindex_hash->{n
} += 1;
8379 # Generate default device list for 'boot: order=' property. Matches legacy
8380 # default boot order, but with explicit device names. This is important, since
8381 # the fallback for when neither 'order' nor the old format is specified relies
8382 # on 'bootorder_from_legacy' above, and it would be confusing if this diverges.
8383 sub get_default_bootdevices
{
8389 my $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 0);
8390 push @ret, $first if $first;
8393 $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 1);
8394 push @ret, $first if $first;
8397 for (my $i = 0; $i < $MAX_NETS; $i++) {
8398 my $netname = "net$i";
8399 next if !$conf->{$netname};
8400 push @ret, $netname;
8407 sub device_bootorder
{
8410 return bootorder_from_legacy
($conf) if !defined($conf->{boot
});
8412 my $boot = parse_property_string
($boot_fmt, $conf->{boot
});
8415 if (!defined($boot) || $boot->{legacy
}) {
8416 $bootorder = bootorder_from_legacy
($conf, $boot);
8417 } elsif ($boot->{order
}) {
8418 my $i = 100; # start at 100 to allow user to insert devices before us with -args
8419 for my $dev (PVE
::Tools
::split_list
($boot->{order
})) {
8420 $bootorder->{$dev} = $i++;
8427 sub register_qmeventd_handle
{
8431 my $peer = "/var/run/qmeventd.sock";
8436 $fh = IO
::Socket
::UNIX-
>new(Peer
=> $peer, Blocking
=> 0, Timeout
=> 1);
8438 if ($! != EINTR
&& $! != EAGAIN
) {
8439 die "unable to connect to qmeventd socket (vmid: $vmid) - $!\n";
8442 die "unable to connect to qmeventd socket (vmid: $vmid) - timeout "
8443 . "after $count retries\n";
8448 # send handshake to mark VM as backing up
8449 print $fh to_json
({vzdump
=> {vmid
=> "$vmid"}});
8451 # return handle to be closed later when inhibit is no longer required
8455 # bash completion helper
8457 sub complete_backup_archives
{
8458 my ($cmdname, $pname, $cvalue) = @_;
8460 my $cfg = PVE
::Storage
::config
();
8464 if ($cvalue =~ m/^([^:]+):/) {
8468 my $data = PVE
::Storage
::template_list
($cfg, $storeid, 'backup');
8471 foreach my $id (keys %$data) {
8472 foreach my $item (@{$data->{$id}}) {
8473 next if $item->{format
} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
8474 push @$res, $item->{volid
} if defined($item->{volid
});
8481 my $complete_vmid_full = sub {
8484 my $idlist = vmstatus
();
8488 foreach my $id (keys %$idlist) {
8489 my $d = $idlist->{$id};
8490 if (defined($running)) {
8491 next if $d->{template
};
8492 next if $running && $d->{status
} ne 'running';
8493 next if !$running && $d->{status
} eq 'running';
8502 return &$complete_vmid_full();
8505 sub complete_vmid_stopped
{
8506 return &$complete_vmid_full(0);
8509 sub complete_vmid_running
{
8510 return &$complete_vmid_full(1);
8513 sub complete_storage
{
8515 my $cfg = PVE
::Storage
::config
();
8516 my $ids = $cfg->{ids
};
8519 foreach my $sid (keys %$ids) {
8520 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, undef, 1);
8521 next if !$ids->{$sid}->{content
}->{images
};
8528 sub complete_migration_storage
{
8529 my ($cmd, $param, $current_value, $all_args) = @_;
8531 my $targetnode = @$all_args[1];
8533 my $cfg = PVE
::Storage
::config
();
8534 my $ids = $cfg->{ids
};
8537 foreach my $sid (keys %$ids) {
8538 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, $targetnode, 1);
8539 next if !$ids->{$sid}->{content
}->{images
};
8548 my $qmpstatus = eval {
8549 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid);
8550 mon_cmd
($vmid, "query-status");
8553 return $qmpstatus && $qmpstatus->{status
} eq "paused";
8556 sub check_volume_storage_type
{
8557 my ($storecfg, $vol) = @_;
8559 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($vol);
8560 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8561 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $vol);
8563 die "storage '$storeid' does not support content-type '$vtype'\n"
8564 if !$scfg->{content
}->{$vtype};
8569 sub add_nets_bridge_fdb
{
8570 my ($conf, $vmid) = @_;
8572 for my $opt (keys %$conf) {
8573 next if $opt !~ m/^net(\d+)$/;
8574 my $iface = "tap${vmid}i$1";
8575 # NOTE: expect setups with learning off to *not* use auto-random-generation of MAC on start
8576 my $net = parse_net
($conf->{$opt}, 1) or next;
8578 my $mac = $net->{macaddr
};
8580 log_warn
("MAC learning disabled, but vNIC '$iface' has no static MAC to add to forwarding DB!")
8581 if !file_read_firstline
("/sys/class/net/$iface/brport/learning");
8585 my $bridge = $net->{bridge
};
8587 log_warn
("Interface '$iface' not attached to any bridge.");
8591 PVE
::Network
::SDN
::Zones
::add_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8592 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8593 PVE
::Network
::add_bridge_fdb
($iface, $mac, $net->{firewall
});
8598 sub del_nets_bridge_fdb
{
8599 my ($conf, $vmid) = @_;
8601 for my $opt (keys %$conf) {
8602 next if $opt !~ m/^net(\d+)$/;
8603 my $iface = "tap${vmid}i$1";
8605 my $net = parse_net
($conf->{$opt}) or next;
8606 my $mac = $net->{macaddr
} or next;
8608 my $bridge = $net->{bridge
};
8610 PVE
::Network
::SDN
::Zones
::del_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8611 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8612 PVE
::Network
::del_bridge_fdb
($iface, $mac, $net->{firewall
});