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 -no-hpet
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 $confdesc_cloudinit = {
765 description
=> 'Specifies the cloud-init configuration format. The default depends on the'
766 .' configured operating system type (`ostype`. We use the `nocloud` format for Linux,'
767 .' and `configdrive2` for windows.',
768 enum
=> ['configdrive2', 'nocloud', 'opennebula'],
773 description
=> "cloud-init: User name to change ssh keys and password for instead of the"
774 ." image's configured default user.",
779 description
=> 'cloud-init: Password to assign the user. Using this is generally not'
780 .' recommended. Use ssh keys instead. Also note that older cloud-init versions do not'
781 .' support hashed passwords.',
786 description
=> 'cloud-init: Specify custom files to replace the automatically generated'
788 format
=> 'pve-qm-cicustom',
793 description
=> 'cloud-init: Sets DNS search domains for a container. Create will'
794 .' automatically use the setting from the host if neither searchdomain nor nameserver'
799 type
=> 'string', format
=> 'address-list',
800 description
=> 'cloud-init: Sets DNS server IP address for a container. Create will'
801 .' automatically use the setting from the host if neither searchdomain nor nameserver'
807 format
=> 'urlencoded',
808 description
=> "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
812 # what about other qemu settings ?
814 #machine => 'string',
827 ##soundhw => 'string',
829 while (my ($k, $v) = each %$confdesc) {
830 PVE
::JSONSchema
::register_standard_option
("pve-qm-$k", $v);
833 my $MAX_USB_DEVICES = 14;
835 my $MAX_SERIAL_PORTS = 4;
836 my $MAX_PARALLEL_PORTS = 3;
842 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
843 description
=> "CPUs accessing this NUMA node.",
844 format_description
=> "id[-id];...",
848 description
=> "Amount of memory this NUMA node provides.",
853 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
854 description
=> "Host NUMA nodes to use.",
855 format_description
=> "id[-id];...",
860 enum
=> [qw(preferred bind interleave)],
861 description
=> "NUMA allocation policy.",
865 PVE
::JSONSchema
::register_format
('pve-qm-numanode', $numa_fmt);
868 type
=> 'string', format
=> $numa_fmt,
869 description
=> "NUMA topology.",
871 PVE
::JSONSchema
::register_standard_option
("pve-qm-numanode", $numadesc);
873 for (my $i = 0; $i < $MAX_NUMA; $i++) {
874 $confdesc->{"numa$i"} = $numadesc;
877 my $nic_model_list = [
893 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
895 my $net_fmt_bridge_descr = <<__EOD__;
896 Bridge to attach the network device to. The Proxmox VE standard bridge
899 If you do not specify a bridge, we create a kvm user (NATed) network
900 device, which provides DHCP and DNS services. The following addresses
907 The DHCP server assign addresses to the guest starting from 10.0.2.15.
911 macaddr
=> get_standard_option
('mac-addr', {
912 description
=> "MAC address. That address must be unique withing your network. This is"
913 ." automatically generated if not specified.",
917 description
=> "Network Card Model. The 'virtio' model provides the best performance with"
918 ." very low CPU overhead. If your guest does not support this driver, it is usually"
919 ." best to use 'e1000'.",
920 enum
=> $nic_model_list,
923 (map { $_ => { keyAlias
=> 'model', alias
=> 'macaddr' }} @$nic_model_list),
924 bridge
=> get_standard_option
('pve-bridge-id', {
925 description
=> $net_fmt_bridge_descr,
930 minimum
=> 0, maximum
=> 64,
931 description
=> 'Number of packet queues to be used on the device.',
937 description
=> "Rate limit in mbps (megabytes per second) as floating point number.",
942 minimum
=> 1, maximum
=> 4094,
943 description
=> 'VLAN tag to apply to packets on this interface.',
948 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
949 description
=> 'VLAN trunks to pass through this interface.',
950 format_description
=> 'vlanid[;vlanid...]',
955 description
=> 'Whether this interface should be protected by the firewall.',
960 description
=> 'Whether this interface should be disconnected (like pulling the plug).',
965 minimum
=> 1, maximum
=> 65520,
966 description
=> "Force MTU, for VirtIO only. Set to '1' to use the bridge MTU",
973 type
=> 'string', format
=> $net_fmt,
974 description
=> "Specify network devices.",
977 PVE
::JSONSchema
::register_standard_option
("pve-qm-net", $netdesc);
982 format
=> 'pve-ipv4-config',
983 format_description
=> 'IPv4Format/CIDR',
984 description
=> 'IPv4 address in CIDR format.',
991 format_description
=> 'GatewayIPv4',
992 description
=> 'Default gateway for IPv4 traffic.',
998 format
=> 'pve-ipv6-config',
999 format_description
=> 'IPv6Format/CIDR',
1000 description
=> 'IPv6 address in CIDR format.',
1007 format_description
=> 'GatewayIPv6',
1008 description
=> 'Default gateway for IPv6 traffic.',
1013 PVE
::JSONSchema
::register_format
('pve-qm-ipconfig', $ipconfig_fmt);
1014 my $ipconfigdesc = {
1016 type
=> 'string', format
=> 'pve-qm-ipconfig',
1017 description
=> <<'EODESCR',
1018 cloud-init: Specify IP addresses and gateways for the corresponding interface.
1020 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
1022 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit
1023 gateway should be provided.
1024 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration. This requires
1025 cloud-init 19.4 or newer.
1027 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using
1031 PVE
::JSONSchema
::register_standard_option
("pve-qm-ipconfig", $netdesc);
1033 for (my $i = 0; $i < $MAX_NETS; $i++) {
1034 $confdesc->{"net$i"} = $netdesc;
1035 $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
1038 foreach my $key (keys %$confdesc_cloudinit) {
1039 $confdesc->{$key} = $confdesc_cloudinit->{$key};
1042 PVE
::JSONSchema
::register_format
('pve-cpuset', \
&pve_verify_cpuset
);
1043 sub pve_verify_cpuset
{
1044 my ($set_text, $noerr) = @_;
1046 my ($count, $members) = eval { PVE
::CpuSet
::parse_cpuset
($set_text) };
1050 die "unable to parse cpuset option\n";
1053 return PVE
::CpuSet-
>new($members)->short_string();
1056 PVE
::JSONSchema
::register_format
('pve-volume-id-or-qm-path', \
&verify_volume_id_or_qm_path
);
1057 sub verify_volume_id_or_qm_path
{
1058 my ($volid, $noerr) = @_;
1060 return $volid if $volid eq 'none' || $volid eq 'cdrom';
1062 return verify_volume_id_or_absolute_path
($volid, $noerr);
1065 PVE
::JSONSchema
::register_format
('pve-volume-id-or-absolute-path', \
&verify_volume_id_or_absolute_path
);
1066 sub verify_volume_id_or_absolute_path
{
1067 my ($volid, $noerr) = @_;
1069 return $volid if $volid =~ m
|^/|;
1071 $volid = eval { PVE
::JSONSchema
::check_format
('pve-volume-id', $volid, '') };
1082 type
=> 'string', format
=> 'pve-qm-usb-device',
1083 format_description
=> 'HOSTUSBDEVICE|spice',
1084 description
=> <<EODESCR,
1085 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1087 'bus-port(.port)*' (decimal numbers) or
1088 'vendor_id:product_id' (hexadeciaml numbers) or
1091 You can use the 'lsusb -t' command to list existing usb devices.
1093 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1094 machines - use with special care.
1096 The value 'spice' can be used to add a usb redirection devices for spice.
1102 description
=> "Specifies whether if given host option is a USB3 device or port."
1103 ." For modern guests (machine version >= 7.1 and ostype l26 and windows > 7), this flag"
1104 ." is irrelevant (all devices are plugged into a xhci controller).",
1111 type
=> 'string', format
=> $usb_fmt,
1112 description
=> "Configure an USB device (n is 0 to 4, for machine version >= 7.1 and ostype"
1113 ." l26 or windows > 7, n can be up to 14).",
1115 PVE
::JSONSchema
::register_standard_option
("pve-qm-usb", $usbdesc);
1120 pattern
=> '(/dev/.+|socket)',
1121 description
=> "Create a serial device inside the VM (n is 0 to 3)",
1122 verbose_description
=> <<EODESCR,
1123 Create a serial device inside the VM (n is 0 to 3), and pass through a
1124 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1125 host side (use 'qm terminal' to open a terminal connection).
1127 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines -
1128 use with special care.
1130 CAUTION: Experimental! User reported problems with this option.
1137 pattern
=> '/dev/parport\d+|/dev/usb/lp\d+',
1138 description
=> "Map host parallel devices (n is 0 to 2).",
1139 verbose_description
=> <<EODESCR,
1140 Map host parallel devices (n is 0 to 2).
1142 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1143 machines - use with special care.
1145 CAUTION: Experimental! User reported problems with this option.
1149 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
1150 $confdesc->{"parallel$i"} = $paralleldesc;
1153 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
1154 $confdesc->{"serial$i"} = $serialdesc;
1157 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
1158 $confdesc->{"hostpci$i"} = $PVE::QemuServer
::PCI
::hostpcidesc
;
1161 for my $key (keys %{$PVE::QemuServer
::Drive
::drivedesc_hash
}) {
1162 $confdesc->{$key} = $PVE::QemuServer
::Drive
::drivedesc_hash-
>{$key};
1165 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
1166 $confdesc->{"usb$i"} = $usbdesc;
1174 description
=> "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n)."
1175 . " Deprecated, use 'order=' instead.",
1176 pattern
=> '[acdn]{1,4}',
1177 format_description
=> "[acdn]{1,4}",
1179 # note: this is also the fallback if boot: is not given at all
1185 format
=> 'pve-qm-bootdev-list',
1186 format_description
=> "device[;device...]",
1187 description
=> <<EODESC,
1188 The guest will attempt to boot from devices in the order they appear here.
1190 Disks, optical drives and passed-through storage USB devices will be directly
1191 booted from, NICs will load PXE, and PCIe devices will either behave like disks
1192 (e.g. NVMe) or load an option ROM (e.g. RAID controller, hardware NIC).
1194 Note that only devices in this list will be marked as bootable and thus loaded
1195 by the guest firmware (BIOS/UEFI). If you require multiple disks for booting
1196 (e.g. software-raid), you need to specify all of them here.
1198 Overrides the deprecated 'legacy=[acdn]*' value when given.
1202 PVE
::JSONSchema
::register_format
('pve-qm-boot', $boot_fmt);
1204 PVE
::JSONSchema
::register_format
('pve-qm-bootdev', \
&verify_bootdev
);
1205 sub verify_bootdev
{
1206 my ($dev, $noerr) = @_;
1208 my $special = $dev =~ m/^efidisk/ || $dev =~ m/^tpmstate/;
1209 return $dev if PVE
::QemuServer
::Drive
::is_valid_drivename
($dev) && !$special;
1213 return 0 if $dev !~ m/^$base\d+$/;
1214 return 0 if !$confdesc->{$dev};
1218 return $dev if $check->("net");
1219 return $dev if $check->("usb");
1220 return $dev if $check->("hostpci");
1223 die "invalid boot device '$dev'\n";
1226 sub print_bootorder
{
1228 return "" if !@$devs;
1229 my $data = { order
=> join(';', @$devs) };
1230 return PVE
::JSONSchema
::print_property_string
($data, $boot_fmt);
1233 my $kvm_api_version = 0;
1236 return $kvm_api_version if $kvm_api_version;
1238 open my $fh, '<', '/dev/kvm' or return;
1240 # 0xae00 => KVM_GET_API_VERSION
1241 $kvm_api_version = ioctl($fh, 0xae00, 0);
1244 return $kvm_api_version;
1247 my $kvm_user_version = {};
1250 sub kvm_user_version
{
1253 $binary //= get_command_for_arch
(get_host_arch
()); # get the native arch by default
1254 my $st = stat($binary);
1256 my $cachedmtime = $kvm_mtime->{$binary} // -1;
1257 return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1258 $cachedmtime == $st->mtime;
1260 $kvm_user_version->{$binary} = 'unknown';
1261 $kvm_mtime->{$binary} = $st->mtime;
1265 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1266 $kvm_user_version->{$binary} = $2;
1270 eval { run_command
([$binary, '--version'], outfunc
=> $code); };
1273 return $kvm_user_version->{$binary};
1276 my sub extract_version
{
1277 my ($machine_type, $version) = @_;
1278 $version = kvm_user_version
() if !defined($version);
1279 return PVE
::QemuServer
::Machine
::extract_version
($machine_type, $version)
1282 sub kernel_has_vhost_net
{
1283 return -c
'/dev/vhost-net';
1288 return defined($confdesc->{$key});
1292 sub get_cdrom_path
{
1294 return $cdrom_path if defined($cdrom_path);
1296 $cdrom_path = first
{ -l
$_ } map { "/dev/cdrom$_" } ('', '1', '2');
1298 if (!defined($cdrom_path)) {
1299 log_warn
("no physical CD-ROM available, ignoring");
1307 my ($storecfg, $vmid, $cdrom) = @_;
1309 if ($cdrom eq 'cdrom') {
1310 return get_cdrom_path
();
1311 } elsif ($cdrom eq 'none') {
1313 } elsif ($cdrom =~ m
|^/|) {
1316 return PVE
::Storage
::path
($storecfg, $cdrom);
1320 # try to convert old style file names to volume IDs
1321 sub filename_to_volume_id
{
1322 my ($vmid, $file, $media) = @_;
1324 if (!($file eq 'none' || $file eq 'cdrom' ||
1325 $file =~ m
|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1327 return if $file =~ m
|/|;
1329 if ($media && $media eq 'cdrom') {
1330 $file = "local:iso/$file";
1332 $file = "local:$vmid/$file";
1339 sub verify_media_type
{
1340 my ($opt, $vtype, $media) = @_;
1345 if ($media eq 'disk') {
1347 } elsif ($media eq 'cdrom') {
1350 die "internal error";
1353 return if ($vtype eq $etype);
1355 raise_param_exc
({ $opt => "unexpected media type ($vtype != $etype)" });
1358 sub cleanup_drive_path
{
1359 my ($opt, $storecfg, $drive) = @_;
1361 # try to convert filesystem paths to volume IDs
1363 if (($drive->{file
} !~ m/^(cdrom|none)$/) &&
1364 ($drive->{file
} !~ m
|^/dev/.+|) &&
1365 ($drive->{file
} !~ m/^([^:]+):(.+)$/) &&
1366 ($drive->{file
} !~ m/^\d+$/)) {
1367 my ($vtype, $volid) = PVE
::Storage
::path_to_volume_id
($storecfg, $drive->{file
});
1368 raise_param_exc
({ $opt => "unable to associate path '$drive->{file}' to any storage"})
1370 $drive->{media
} = 'cdrom' if !$drive->{media
} && $vtype eq 'iso';
1371 verify_media_type
($opt, $vtype, $drive->{media
});
1372 $drive->{file
} = $volid;
1375 $drive->{media
} = 'cdrom' if !$drive->{media
} && $drive->{file
} =~ m/^(cdrom|none)$/;
1378 sub parse_hotplug_features
{
1383 return $res if $data eq '0';
1385 $data = $confdesc->{hotplug
}->{default} if $data eq '1';
1387 foreach my $feature (PVE
::Tools
::split_list
($data)) {
1388 if ($feature =~ m/^(network|disk|cpu|memory|usb|cloudinit)$/) {
1391 die "invalid hotplug feature '$feature'\n";
1397 PVE
::JSONSchema
::register_format
('pve-hotplug-features', \
&pve_verify_hotplug_features
);
1398 sub pve_verify_hotplug_features
{
1399 my ($value, $noerr) = @_;
1401 return $value if parse_hotplug_features
($value);
1405 die "unable to parse hotplug option\n";
1409 my($fh, $noerr) = @_;
1412 my $SG_GET_VERSION_NUM = 0x2282;
1414 my $versionbuf = "\x00" x
8;
1415 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1417 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1420 my $version = unpack("I", $versionbuf);
1421 if ($version < 30000) {
1422 die "scsi generic interface too old\n" if !$noerr;
1426 my $buf = "\x00" x
36;
1427 my $sensebuf = "\x00" x
8;
1428 my $cmd = pack("C x3 C x1", 0x12, 36);
1430 # see /usr/include/scsi/sg.h
1431 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";
1434 $sg_io_hdr_t, ord('S'), -3, length($cmd), length($sensebuf), 0, length($buf), $buf, $cmd, $sensebuf, 6000
1437 $ret = ioctl($fh, $SG_IO, $packet);
1439 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1443 my @res = unpack($sg_io_hdr_t, $packet);
1444 if ($res[17] || $res[18]) {
1445 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1450 $res->@{qw(type removable vendor product revision)} = unpack("C C x6 A8 A16 A4", $buf);
1452 $res->{removable
} = $res->{removable
} & 128 ?
1 : 0;
1453 $res->{type
} &= 0x1F;
1461 my $fh = IO
::File-
>new("+<$path") || return;
1462 my $res = scsi_inquiry
($fh, 1);
1468 sub print_tabletdevice_full
{
1469 my ($conf, $arch) = @_;
1471 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1473 # we use uhci for old VMs because tablet driver was buggy in older qemu
1475 if ($q35 || $arch eq 'aarch64') {
1481 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1484 sub print_keyboarddevice_full
{
1485 my ($conf, $arch) = @_;
1487 return if $arch ne 'aarch64';
1489 return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1492 my sub get_drive_id
{
1494 return "$drive->{interface}$drive->{index}";
1497 sub print_drivedevice_full
{
1498 my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1503 my $drive_id = get_drive_id
($drive);
1504 if ($drive->{interface
} eq 'virtio') {
1505 my $pciaddr = print_pci_addr
("$drive_id", $bridges, $arch, $machine_type);
1506 $device = "virtio-blk-pci,drive=drive-$drive_id,id=${drive_id}${pciaddr}";
1507 $device .= ",iothread=iothread-$drive_id" if $drive->{iothread
};
1508 } elsif ($drive->{interface
} eq 'scsi') {
1510 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
1511 my $unit = $drive->{index} % $maxdev;
1512 my $devicetype = 'hd';
1514 if (drive_is_cdrom
($drive)) {
1517 if ($drive->{file
} =~ m
|^/|) {
1518 $path = $drive->{file
};
1519 if (my $info = path_is_scsi
($path)) {
1520 if ($info->{type
} == 0 && $drive->{scsiblock
}) {
1521 $devicetype = 'block';
1522 } elsif ($info->{type
} == 1) { # tape
1523 $devicetype = 'generic';
1527 $path = PVE
::Storage
::path
($storecfg, $drive->{file
});
1530 # for compatibility only, we prefer scsi-hd (#2408, #2355, #2380)
1531 my $version = extract_version
($machine_type, kvm_user_version
());
1532 if ($path =~ m/^iscsi\:\/\
// &&
1533 !min_version
($version, 4, 1)) {
1534 $devicetype = 'generic';
1538 if (!$conf->{scsihw
} || $conf->{scsihw
} =~ m/^lsi/ || $conf->{scsihw
} eq 'pvscsi') {
1539 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit";
1541 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0"
1542 .",lun=$drive->{index}";
1544 $device .= ",drive=drive-$drive_id,id=$drive_id";
1546 if ($drive->{ssd
} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1547 $device .= ",rotation_rate=1";
1549 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1551 } elsif ($drive->{interface
} eq 'ide' || $drive->{interface
} eq 'sata') {
1552 my $maxdev = ($drive->{interface
} eq 'sata') ?
$PVE::QemuServer
::Drive
::MAX_SATA_DISKS
: 2;
1553 my $controller = int($drive->{index} / $maxdev);
1554 my $unit = $drive->{index} % $maxdev;
1555 my $devicetype = ($drive->{media
} && $drive->{media
} eq 'cdrom') ?
"cd" : "hd";
1557 $device = "ide-$devicetype";
1558 if ($drive->{interface
} eq 'ide') {
1559 $device .= ",bus=ide.$controller,unit=$unit";
1561 $device .= ",bus=ahci$controller.$unit";
1563 $device .= ",drive=drive-$drive_id,id=$drive_id";
1565 if ($devicetype eq 'hd') {
1566 if (my $model = $drive->{model
}) {
1567 $model = URI
::Escape
::uri_unescape
($model);
1568 $device .= ",model=$model";
1570 if ($drive->{ssd
}) {
1571 $device .= ",rotation_rate=1";
1574 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1575 } elsif ($drive->{interface
} eq 'usb') {
1577 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1579 die "unsupported interface type";
1582 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex
};
1584 if (my $serial = $drive->{serial
}) {
1585 $serial = URI
::Escape
::uri_unescape
($serial);
1586 $device .= ",serial=$serial";
1593 sub get_initiator_name
{
1596 my $fh = IO
::File-
>new('/etc/iscsi/initiatorname.iscsi') || return;
1597 while (defined(my $line = <$fh>)) {
1598 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1607 my sub storage_allows_io_uring_default
{
1608 my ($scfg, $cache_direct) = @_;
1610 # io_uring with cache mode writeback or writethrough on krbd will hang...
1611 return if $scfg && $scfg->{type
} eq 'rbd' && $scfg->{krbd
} && !$cache_direct;
1613 # io_uring with cache mode writeback or writethrough on LVM will hang, without cache only
1614 # sometimes, just plain disable...
1615 return if $scfg && $scfg->{type
} eq 'lvm';
1617 # io_uring causes problems when used with CIFS since kernel 5.15
1618 # Some discussion: https://www.spinics.net/lists/linux-cifs/msg26734.html
1619 return if $scfg && $scfg->{type
} eq 'cifs';
1624 my sub drive_uses_cache_direct
{
1625 my ($drive, $scfg) = @_;
1627 my $cache_direct = 0;
1629 if (my $cache = $drive->{cache
}) {
1630 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1631 } elsif (!drive_is_cdrom
($drive) && !($scfg && $scfg->{type
} eq 'btrfs' && !$scfg->{nocow
})) {
1635 return $cache_direct;
1638 sub print_drive_commandline_full
{
1639 my ($storecfg, $vmid, $drive, $pbs_name, $io_uring) = @_;
1642 my $volid = $drive->{file
};
1643 my $format = $drive->{format
};
1644 my $drive_id = get_drive_id
($drive);
1646 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1647 my $scfg = $storeid ? PVE
::Storage
::storage_config
($storecfg, $storeid) : undef;
1649 if (drive_is_cdrom
($drive)) {
1650 $path = get_iso_path
($storecfg, $vmid, $volid);
1651 die "$drive_id: cannot back cdrom drive with PBS snapshot\n" if $pbs_name;
1654 $path = PVE
::Storage
::path
($storecfg, $volid);
1655 $format //= qemu_img_format
($scfg, $volname);
1662 my $is_rbd = $path =~ m/^rbd:/;
1665 my @qemu_drive_options = qw(heads secs cyls trans media cache rerror werror aio discard);
1666 foreach my $o (@qemu_drive_options) {
1667 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1670 # snapshot only accepts on|off
1671 if (defined($drive->{snapshot
})) {
1672 my $v = $drive->{snapshot
} ?
'on' : 'off';
1673 $opts .= ",snapshot=$v";
1676 if (defined($drive->{ro
})) { # ro maps to QEMUs `readonly`, which accepts `on` or `off` only
1677 $opts .= ",readonly=" . ($drive->{ro
} ?
'on' : 'off');
1680 foreach my $type (['', '-total'], [_rd
=> '-read'], [_wr
=> '-write']) {
1681 my ($dir, $qmpname) = @$type;
1682 if (my $v = $drive->{"mbps$dir"}) {
1683 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1685 if (my $v = $drive->{"mbps${dir}_max"}) {
1686 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1688 if (my $v = $drive->{"bps${dir}_max_length"}) {
1689 $opts .= ",throttling.bps$qmpname-max-length=$v";
1691 if (my $v = $drive->{"iops${dir}"}) {
1692 $opts .= ",throttling.iops$qmpname=$v";
1694 if (my $v = $drive->{"iops${dir}_max"}) {
1695 $opts .= ",throttling.iops$qmpname-max=$v";
1697 if (my $v = $drive->{"iops${dir}_max_length"}) {
1698 $opts .= ",throttling.iops$qmpname-max-length=$v";
1703 $format = "rbd" if $is_rbd;
1704 die "$drive_id: Proxmox Backup Server backed drive cannot auto-detect the format\n"
1706 $opts .= ",format=alloc-track,file.driver=$format";
1708 $opts .= ",format=$format";
1711 my $cache_direct = drive_uses_cache_direct
($drive, $scfg);
1713 $opts .= ",cache=none" if !$drive->{cache
} && $cache_direct;
1715 if (!$drive->{aio
}) {
1716 if ($io_uring && storage_allows_io_uring_default
($scfg, $cache_direct)) {
1717 # io_uring supports all cache modes
1718 $opts .= ",aio=io_uring";
1720 # aio native works only with O_DIRECT
1722 $opts .= ",aio=native";
1724 $opts .= ",aio=threads";
1729 if (!drive_is_cdrom
($drive)) {
1731 if (defined($drive->{detect_zeroes
}) && !$drive->{detect_zeroes
}) {
1732 $detectzeroes = 'off';
1733 } elsif ($drive->{discard
}) {
1734 $detectzeroes = $drive->{discard
} eq 'on' ?
'unmap' : 'on';
1736 # This used to be our default with discard not being specified:
1737 $detectzeroes = 'on';
1740 # note: 'detect-zeroes' works per blockdev and we want it to persist
1741 # after the alloc-track is removed, so put it on 'file' directly
1742 my $dz_param = $pbs_name ?
"file.detect-zeroes" : "detect-zeroes";
1743 $opts .= ",$dz_param=$detectzeroes" if $detectzeroes;
1747 $opts .= ",backing=$pbs_name";
1748 $opts .= ",auto-remove=on";
1751 # my $file_param = $pbs_name ? "file.file.filename" : "file";
1752 my $file_param = "file";
1754 # non-rbd drivers require the underlying file to be a seperate block
1755 # node, so add a second .file indirection
1756 $file_param .= ".file" if !$is_rbd;
1757 $file_param .= ".filename";
1759 my $pathinfo = $path ?
"$file_param=$path," : '';
1761 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1764 sub print_pbs_blockdev
{
1765 my ($pbs_conf, $pbs_name) = @_;
1766 my $blockdev = "driver=pbs,node-name=$pbs_name,read-only=on";
1767 $blockdev .= ",repository=$pbs_conf->{repository}";
1768 $blockdev .= ",namespace=$pbs_conf->{namespace}" if $pbs_conf->{namespace
};
1769 $blockdev .= ",snapshot=$pbs_conf->{snapshot}";
1770 $blockdev .= ",archive=$pbs_conf->{archive}";
1771 $blockdev .= ",keyfile=$pbs_conf->{keyfile}" if $pbs_conf->{keyfile
};
1775 sub print_netdevice_full
{
1776 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version) = @_;
1778 my $device = $net->{model
};
1779 if ($net->{model
} eq 'virtio') {
1780 $device = 'virtio-net-pci';
1783 my $pciaddr = print_pci_addr
("$netid", $bridges, $arch, $machine_type);
1784 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1785 if ($net->{queues
} && $net->{queues
} > 1 && $net->{model
} eq 'virtio'){
1786 # Consider we have N queues, the number of vectors needed is 2 * N + 2, i.e., one per in
1787 # and out of each queue plus one config interrupt and control vector queue
1788 my $vectors = $net->{queues
} * 2 + 2;
1789 $tmpstr .= ",vectors=$vectors,mq=on";
1790 if (min_version
($machine_version, 7, 1)) {
1791 $tmpstr .= ",packed=on";
1795 if (min_version
($machine_version, 7, 1) && $net->{model
} eq 'virtio'){
1796 $tmpstr .= ",rx_queue_size=1024,tx_queue_size=1024";
1799 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex
} ;
1801 if (my $mtu = $net->{mtu
}) {
1802 if ($net->{model
} eq 'virtio' && $net->{bridge
}) {
1803 my $bridge_mtu = PVE
::Network
::read_bridge_mtu
($net->{bridge
});
1806 } elsif ($mtu < 576) {
1807 die "netdev $netid: MTU '$mtu' is smaller than the IP minimum MTU '576'\n";
1808 } elsif ($mtu > $bridge_mtu) {
1809 die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
1811 $tmpstr .= ",host_mtu=$mtu";
1813 warn "WARN: netdev $netid: ignoring MTU '$mtu', not using VirtIO or no bridge configured.\n";
1817 if ($use_old_bios_files) {
1819 if ($device eq 'virtio-net-pci') {
1820 $romfile = 'pxe-virtio.rom';
1821 } elsif ($device eq 'e1000') {
1822 $romfile = 'pxe-e1000.rom';
1823 } elsif ($device eq 'e1000e') {
1824 $romfile = 'pxe-e1000e.rom';
1825 } elsif ($device eq 'ne2k') {
1826 $romfile = 'pxe-ne2k_pci.rom';
1827 } elsif ($device eq 'pcnet') {
1828 $romfile = 'pxe-pcnet.rom';
1829 } elsif ($device eq 'rtl8139') {
1830 $romfile = 'pxe-rtl8139.rom';
1832 $tmpstr .= ",romfile=$romfile" if $romfile;
1838 sub print_netdev_full
{
1839 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
1842 if ($netid =~ m/^net(\d+)$/) {
1846 die "got strange net id '$i'\n" if $i >= ${MAX_NETS
};
1848 my $ifname = "tap${vmid}i$i";
1850 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1851 die "interface name '$ifname' is too long (max 15 character)\n"
1852 if length($ifname) >= 16;
1854 my $vhostparam = '';
1855 if (is_native
($arch)) {
1856 $vhostparam = ',vhost=on' if kernel_has_vhost_net
() && $net->{model
} eq 'virtio';
1859 my $vmname = $conf->{name
} || "vm$vmid";
1862 my $script = $hotplug ?
"pve-bridge-hotplug" : "pve-bridge";
1864 if ($net->{bridge
}) {
1865 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script"
1866 .",downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1868 $netdev = "type=user,id=$netid,hostname=$vmname";
1871 $netdev .= ",queues=$net->{queues}" if ($net->{queues
} && $net->{model
} eq 'virtio');
1877 'cirrus' => 'cirrus-vga',
1879 'vmware' => 'vmware-svga',
1880 'virtio' => 'virtio-vga',
1881 'virtio-gl' => 'virtio-vga-gl',
1884 sub print_vga_device
{
1885 my ($conf, $vga, $arch, $machine_version, $machine, $id, $qxlnum, $bridges) = @_;
1887 my $type = $vga_map->{$vga->{type
}};
1888 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
1889 $type = 'virtio-gpu';
1891 my $vgamem_mb = $vga->{memory
};
1893 my $max_outputs = '';
1895 $type = $id ?
'qxl' : 'qxl-vga';
1897 if (!$conf->{ostype
} || $conf->{ostype
} =~ m/^(?:l\d\d)|(?:other)$/) {
1898 # set max outputs so linux can have up to 4 qxl displays with one device
1899 if (min_version
($machine_version, 4, 1)) {
1900 $max_outputs = ",max_outputs=4";
1905 die "no devicetype for $vga->{type}\n" if !$type;
1909 if ($vga->{type
} =~ /^virtio/) {
1910 my $bytes = PVE
::Tools
::convert_size
($vgamem_mb, "mb" => "b");
1911 $memory = ",max_hostmem=$bytes";
1913 # from https://www.spice-space.org/multiple-monitors.html
1914 $memory = ",vgamem_mb=$vga->{memory}";
1915 my $ram = $vgamem_mb * 4;
1916 my $vram = $vgamem_mb * 2;
1917 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
1919 $memory = ",vgamem_mb=$vga->{memory}";
1921 } elsif ($qxlnum && $id) {
1922 $memory = ",ram_size=67108864,vram_size=33554432";
1926 if ($type eq 'VGA' && windows_version
($conf->{ostype
})) {
1927 $edidoff=",edid=off" if (!defined($conf->{bios
}) || $conf->{bios
} ne 'ovmf');
1930 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1931 my $vgaid = "vga" . ($id // '');
1933 if ($q35 && $vgaid eq 'vga') {
1934 # the first display uses pcie.0 bus on q35 machines
1935 $pciaddr = print_pcie_addr
($vgaid, $bridges, $arch, $machine);
1937 $pciaddr = print_pci_addr
($vgaid, $bridges, $arch, $machine);
1940 if ($vga->{type
} eq 'virtio-gl') {
1941 my $base = '/usr/lib/x86_64-linux-gnu/lib';
1942 die "missing libraries for '$vga->{type}' detected! Please install 'libgl1' and 'libegl1'\n"
1943 if !-e
"${base}EGL.so.1" || !-e
"${base}GL.so.1";
1945 die "no DRM render node detected (/dev/dri/renderD*), no GPU? - needed for '$vga->{type}' display\n"
1946 if !PVE
::Tools
::dir_glob_regex
('/dev/dri/', "renderD.*");
1949 return "$type,id=${vgaid}${memory}${max_outputs}${pciaddr}${edidoff}";
1952 sub parse_number_sets
{
1955 foreach my $part (split(/;/, $set)) {
1956 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1957 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1958 push @$res, [ $1, $2 ];
1960 die "invalid range: $part\n";
1969 my $res = parse_property_string
($numa_fmt, $data);
1970 $res->{cpus
} = parse_number_sets
($res->{cpus
}) if defined($res->{cpus
});
1971 $res->{hostnodes
} = parse_number_sets
($res->{hostnodes
}) if defined($res->{hostnodes
});
1975 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1977 my ($data, $disable_mac_autogen) = @_;
1979 my $res = eval { parse_property_string
($net_fmt, $data) };
1984 if (!defined($res->{macaddr
}) && !$disable_mac_autogen) {
1985 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
1986 $res->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
});
1991 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
1992 sub parse_ipconfig
{
1995 my $res = eval { parse_property_string
($ipconfig_fmt, $data) };
2001 if ($res->{gw
} && !$res->{ip
}) {
2002 warn 'gateway specified without specifying an IP address';
2005 if ($res->{gw6
} && !$res->{ip6
}) {
2006 warn 'IPv6 gateway specified without specifying an IPv6 address';
2009 if ($res->{gw
} && $res->{ip
} eq 'dhcp') {
2010 warn 'gateway specified together with DHCP';
2013 if ($res->{gw6
} && $res->{ip6
} !~ /^$IPV6RE/) {
2015 warn "IPv6 gateway specified together with $res->{ip6} address";
2019 if (!$res->{ip
} && !$res->{ip6
}) {
2020 return { ip
=> 'dhcp', ip6
=> 'dhcp' };
2029 return PVE
::JSONSchema
::print_property_string
($net, $net_fmt);
2032 sub add_random_macs
{
2033 my ($settings) = @_;
2035 foreach my $opt (keys %$settings) {
2036 next if $opt !~ m/^net(\d+)$/;
2037 my $net = parse_net
($settings->{$opt});
2039 $settings->{$opt} = print_net
($net);
2043 sub vm_is_volid_owner
{
2044 my ($storecfg, $vmid, $volid) = @_;
2046 if ($volid !~ m
|^/|) {
2048 eval { ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid); };
2049 if ($owner && ($owner == $vmid)) {
2057 sub vmconfig_register_unused_drive
{
2058 my ($storecfg, $vmid, $conf, $drive) = @_;
2060 if (drive_is_cloudinit
($drive)) {
2061 eval { PVE
::Storage
::vdisk_free
($storecfg, $drive->{file
}) };
2063 delete $conf->{cloudinit
};
2064 } elsif (!drive_is_cdrom
($drive)) {
2065 my $volid = $drive->{file
};
2066 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
2067 PVE
::QemuConfig-
>add_unused_volume($conf, $volid, $vmid);
2072 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2076 pattern
=> '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2077 format_description
=> 'UUID',
2078 description
=> "Set SMBIOS1 UUID.",
2083 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2084 format_description
=> 'Base64 encoded string',
2085 description
=> "Set SMBIOS1 version.",
2090 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2091 format_description
=> 'Base64 encoded string',
2092 description
=> "Set SMBIOS1 serial number.",
2097 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2098 format_description
=> 'Base64 encoded string',
2099 description
=> "Set SMBIOS1 manufacturer.",
2104 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2105 format_description
=> 'Base64 encoded string',
2106 description
=> "Set SMBIOS1 product ID.",
2111 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2112 format_description
=> 'Base64 encoded string',
2113 description
=> "Set SMBIOS1 SKU string.",
2118 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2119 format_description
=> 'Base64 encoded string',
2120 description
=> "Set SMBIOS1 family string.",
2125 description
=> 'Flag to indicate that the SMBIOS values are base64 encoded',
2133 my $res = eval { parse_property_string
($smbios1_fmt, $data) };
2140 return PVE
::JSONSchema
::print_property_string
($smbios1, $smbios1_fmt);
2143 PVE
::JSONSchema
::register_format
('pve-qm-smbios1', $smbios1_fmt);
2145 sub parse_watchdog
{
2150 my $res = eval { parse_property_string
($watchdog_fmt, $value) };
2155 sub parse_guest_agent
{
2158 return {} if !defined($conf->{agent
});
2160 my $res = eval { parse_property_string
($agent_fmt, $conf->{agent
}) };
2163 # if the agent is disabled ignore the other potentially set properties
2164 return {} if !$res->{enabled
};
2169 my ($conf, $key) = @_;
2170 return undef if !defined($conf->{agent
});
2172 my $agent = parse_guest_agent
($conf);
2173 return $agent->{$key};
2179 return {} if !$value;
2180 my $res = eval { parse_property_string
($vga_fmt, $value) };
2190 my $res = eval { parse_property_string
($rng_fmt, $value) };
2195 sub parse_meta_info
{
2200 my $res = eval { parse_property_string
($meta_info_fmt, $value) };
2205 sub new_meta_info_string
{
2206 my () = @_; # for now do not allow to override any value
2208 return PVE
::JSONSchema
::print_property_string
(
2210 'creation-qemu' => kvm_user_version
(),
2211 ctime
=> "". int(time()),
2217 sub qemu_created_version_fixups
{
2218 my ($conf, $forcemachine, $kvmver) = @_;
2220 my $meta = parse_meta_info
($conf->{meta
}) // {};
2221 my $forced_vers = PVE
::QemuServer
::Machine
::extract_version
($forcemachine);
2223 # check if we need to apply some handling for VMs that always use the latest machine version but
2224 # had a machine version transition happen that affected HW such that, e.g., an OS config change
2225 # would be required (we do not want to pin machine version for non-windows OS type)
2227 (!defined($conf->{machine
}) || $conf->{machine
} =~ m/^(?:pc|q35|virt)$/) # non-versioned machine
2228 && (!defined($meta->{'creation-qemu'}) || !min_version
($meta->{'creation-qemu'}, 6, 1)) # created before 6.1
2229 && (!$forced_vers || min_version
($forced_vers, 6, 1)) # handle snapshot-rollback/migrations
2230 && min_version
($kvmver, 6, 1) # only need to apply the change since 6.1
2232 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
2233 if ($q35 && $conf->{ostype
} && $conf->{ostype
} eq 'l26') {
2234 # this changed to default-on in Q 6.1 for q35 machines, it will mess with PCI slot view
2235 # and thus with the predictable interface naming of systemd
2236 return ['-global', 'ICH9-LPC.acpi-pci-hotplug-with-bridge-support=off'];
2242 PVE
::JSONSchema
::register_format
('pve-qm-usb-device', \
&verify_usb_device
);
2243 sub verify_usb_device
{
2244 my ($value, $noerr) = @_;
2246 return $value if parse_usb_device
($value);
2250 die "unable to parse usb device\n";
2253 # add JSON properties for create and set function
2254 sub json_config_properties
{
2255 my ($prop, $with_disk_alloc) = @_;
2257 my $skip_json_config_opts = {
2261 runningmachine
=> 1,
2266 foreach my $opt (keys %$confdesc) {
2267 next if $skip_json_config_opts->{$opt};
2269 if ($with_disk_alloc && is_valid_drivename
($opt)) {
2270 $prop->{$opt} = $PVE::QemuServer
::Drive
::drivedesc_hash_with_alloc-
>{$opt};
2272 $prop->{$opt} = $confdesc->{$opt};
2279 # Properties that we can read from an OVF file
2280 sub json_ovf_properties
{
2283 for my $device (PVE
::QemuServer
::Drive
::valid_drive_names
()) {
2284 $prop->{$device} = {
2286 format
=> 'pve-volume-id-or-absolute-path',
2287 description
=> "Disk image that gets imported to $device",
2294 description
=> "The number of CPU cores.",
2299 description
=> "Amount of RAM for the VM in MB.",
2304 description
=> "Name of the VM.",
2311 # return copy of $confdesc_cloudinit to generate documentation
2312 sub cloudinit_config_properties
{
2314 return dclone
($confdesc_cloudinit);
2317 sub cloudinit_pending_properties
{
2319 map { $_ => 1 } keys $confdesc_cloudinit->%*,
2322 $p->{"net$_"} = 1 for 0..($MAX_NETS-1);
2327 my ($key, $value) = @_;
2329 die "unknown setting '$key'\n" if !$confdesc->{$key};
2331 my $type = $confdesc->{$key}->{type
};
2333 if (!defined($value)) {
2334 die "got undefined value\n";
2337 if ($value =~ m/[\n\r]/) {
2338 die "property contains a line feed\n";
2341 if ($type eq 'boolean') {
2342 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2343 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2344 die "type check ('boolean') failed - got '$value'\n";
2345 } elsif ($type eq 'integer') {
2346 return int($1) if $value =~ m/^(\d+)$/;
2347 die "type check ('integer') failed - got '$value'\n";
2348 } elsif ($type eq 'number') {
2349 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2350 die "type check ('number') failed - got '$value'\n";
2351 } elsif ($type eq 'string') {
2352 if (my $fmt = $confdesc->{$key}->{format
}) {
2353 PVE
::JSONSchema
::check_format
($fmt, $value);
2356 $value =~ s/^\"(.*)\"$/$1/;
2359 die "internal error"
2364 my ($storecfg, $vmid, $skiplock, $replacement_conf, $purge_unreferenced) = @_;
2366 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2368 if (!$skiplock && !PVE
::QemuConfig-
>has_lock($conf, 'suspended')) {
2369 PVE
::QemuConfig-
>check_lock($conf);
2372 if ($conf->{template
}) {
2373 # check if any base image is still used by a linked clone
2374 PVE
::QemuConfig-
>foreach_volume_full($conf, { include_unused
=> 1 }, sub {
2375 my ($ds, $drive) = @_;
2376 return if drive_is_cdrom
($drive);
2378 my $volid = $drive->{file
};
2379 return if !$volid || $volid =~ m
|^/|;
2381 die "base volume '$volid' is still in use by linked cloned\n"
2382 if PVE
::Storage
::volume_is_base_and_used
($storecfg, $volid);
2388 my $remove_owned_drive = sub {
2389 my ($ds, $drive) = @_;
2390 return if drive_is_cdrom
($drive, 1);
2392 my $volid = $drive->{file
};
2393 return if !$volid || $volid =~ m
|^/|;
2394 return if $volids->{$volid};
2396 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
2397 return if !$path || !$owner || ($owner != $vmid);
2399 $volids->{$volid} = 1;
2400 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2401 warn "Could not remove disk '$volid', check manually: $@" if $@;
2404 # only remove disks owned by this VM (referenced in the config)
2405 my $include_opts = {
2406 include_unused
=> 1,
2407 extra_keys
=> ['vmstate'],
2409 PVE
::QemuConfig-
>foreach_volume_full($conf, $include_opts, $remove_owned_drive);
2411 for my $snap (values %{$conf->{snapshots
}}) {
2412 next if !defined($snap->{vmstate
});
2413 my $drive = PVE
::QemuConfig-
>parse_volume('vmstate', $snap->{vmstate
}, 1);
2414 next if !defined($drive);
2415 $remove_owned_drive->('vmstate', $drive);
2418 PVE
::QemuConfig-
>foreach_volume_full($conf->{pending
}, $include_opts, $remove_owned_drive);
2420 if ($purge_unreferenced) { # also remove unreferenced disk
2421 my $vmdisks = PVE
::Storage
::vdisk_list
($storecfg, undef, $vmid, undef, 'images');
2422 PVE
::Storage
::foreach_volid
($vmdisks, sub {
2423 my ($volid, $sid, $volname, $d) = @_;
2424 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2429 if (defined $replacement_conf) {
2430 PVE
::QemuConfig-
>write_config($vmid, $replacement_conf);
2432 PVE
::QemuConfig-
>destroy_config($vmid);
2436 sub parse_vm_config
{
2437 my ($filename, $raw, $strict) = @_;
2439 return if !defined($raw);
2442 digest
=> Digest
::SHA
::sha1_hex
($raw),
2448 my $handle_error = sub {
2458 $filename =~ m
|/qemu-server/(\d
+)\
.conf
$|
2459 || die "got strange filename '$filename'";
2465 my $finish_description = sub {
2466 if (defined($descr)) {
2468 $conf->{description
} = $descr;
2474 my @lines = split(/\n/, $raw);
2475 foreach my $line (@lines) {
2476 next if $line =~ m/^\s*$/;
2478 if ($line =~ m/^\[PENDING\]\s*$/i) {
2479 $section = 'pending';
2480 $finish_description->();
2481 $conf = $res->{$section} = {};
2483 } elsif ($line =~ m/^\[special:cloudinit\]\s*$/i) {
2484 $section = 'cloudinit';
2485 $finish_description->();
2486 $conf = $res->{$section} = {};
2489 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2491 $finish_description->();
2492 $conf = $res->{snapshots
}->{$section} = {};
2496 if ($line =~ m/^\#(.*)$/) {
2497 $descr = '' if !defined($descr);
2498 $descr .= PVE
::Tools
::decode_text
($1) . "\n";
2502 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2503 $descr = '' if !defined($descr);
2504 $descr .= PVE
::Tools
::decode_text
($2);
2505 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2506 $conf->{snapstate
} = $1;
2507 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2510 $conf->{$key} = $value;
2511 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2513 if ($section eq 'pending') {
2514 $conf->{delete} = $value; # we parse this later
2516 $handle_error->("vm $vmid - property 'delete' is only allowed in [PENDING]\n");
2518 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2521 if ($section eq 'cloudinit') {
2522 # ignore validation only used for informative purpose
2523 $conf->{$key} = $value;
2526 eval { $value = check_type
($key, $value); };
2528 $handle_error->("vm $vmid - unable to parse value of '$key' - $@");
2530 $key = 'ide2' if $key eq 'cdrom';
2531 my $fmt = $confdesc->{$key}->{format
};
2532 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2533 my $v = parse_drive
($key, $value);
2534 if (my $volid = filename_to_volume_id
($vmid, $v->{file
}, $v->{media
})) {
2535 $v->{file
} = $volid;
2536 $value = print_drive
($v);
2538 $handle_error->("vm $vmid - unable to parse value of '$key'\n");
2543 $conf->{$key} = $value;
2546 $handle_error->("vm $vmid - unable to parse config: $line\n");
2550 $finish_description->();
2551 delete $res->{snapstate
}; # just to be sure
2556 sub write_vm_config
{
2557 my ($filename, $conf) = @_;
2559 delete $conf->{snapstate
}; # just to be sure
2561 if ($conf->{cdrom
}) {
2562 die "option ide2 conflicts with cdrom\n" if $conf->{ide2
};
2563 $conf->{ide2
} = $conf->{cdrom
};
2564 delete $conf->{cdrom
};
2567 # we do not use 'smp' any longer
2568 if ($conf->{sockets
}) {
2569 delete $conf->{smp
};
2570 } elsif ($conf->{smp
}) {
2571 $conf->{sockets
} = $conf->{smp
};
2572 delete $conf->{cores
};
2573 delete $conf->{smp
};
2576 my $used_volids = {};
2578 my $cleanup_config = sub {
2579 my ($cref, $pending, $snapname) = @_;
2581 foreach my $key (keys %$cref) {
2582 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2583 $key eq 'snapstate' || $key eq 'pending' || $key eq 'cloudinit';
2584 my $value = $cref->{$key};
2585 if ($key eq 'delete') {
2586 die "propertry 'delete' is only allowed in [PENDING]\n"
2588 # fixme: check syntax?
2591 eval { $value = check_type
($key, $value); };
2592 die "unable to parse value of '$key' - $@" if $@;
2594 $cref->{$key} = $value;
2596 if (!$snapname && is_valid_drivename
($key)) {
2597 my $drive = parse_drive
($key, $value);
2598 $used_volids->{$drive->{file
}} = 1 if $drive && $drive->{file
};
2603 &$cleanup_config($conf);
2605 &$cleanup_config($conf->{pending
}, 1);
2607 foreach my $snapname (keys %{$conf->{snapshots
}}) {
2608 die "internal error: snapshot name '$snapname' is forbidden" if lc($snapname) eq 'pending';
2609 &$cleanup_config($conf->{snapshots
}->{$snapname}, undef, $snapname);
2612 # remove 'unusedX' settings if we re-add a volume
2613 foreach my $key (keys %$conf) {
2614 my $value = $conf->{$key};
2615 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2616 delete $conf->{$key};
2620 my $generate_raw_config = sub {
2621 my ($conf, $pending) = @_;
2625 # add description as comment to top of file
2626 if (defined(my $descr = $conf->{description
})) {
2628 foreach my $cl (split(/\n/, $descr)) {
2629 $raw .= '#' . PVE
::Tools
::encode_text
($cl) . "\n";
2632 $raw .= "#\n" if $pending;
2636 foreach my $key (sort keys %$conf) {
2637 next if $key =~ /^(digest|description|pending|cloudinit|snapshots)$/;
2638 $raw .= "$key: $conf->{$key}\n";
2643 my $raw = &$generate_raw_config($conf);
2645 if (scalar(keys %{$conf->{pending
}})){
2646 $raw .= "\n[PENDING]\n";
2647 $raw .= &$generate_raw_config($conf->{pending
}, 1);
2650 if (scalar(keys %{$conf->{cloudinit
}}) && PVE
::QemuConfig-
>has_cloudinit($conf)){
2651 $raw .= "\n[special:cloudinit]\n";
2652 $raw .= &$generate_raw_config($conf->{cloudinit
});
2655 foreach my $snapname (sort keys %{$conf->{snapshots
}}) {
2656 $raw .= "\n[$snapname]\n";
2657 $raw .= &$generate_raw_config($conf->{snapshots
}->{$snapname});
2667 # we use static defaults from our JSON schema configuration
2668 foreach my $key (keys %$confdesc) {
2669 if (defined(my $default = $confdesc->{$key}->{default})) {
2670 $res->{$key} = $default;
2678 my $vmlist = PVE
::Cluster
::get_vmlist
();
2680 return $res if !$vmlist || !$vmlist->{ids
};
2681 my $ids = $vmlist->{ids
};
2682 my $nodename = nodename
();
2684 foreach my $vmid (keys %$ids) {
2685 my $d = $ids->{$vmid};
2686 next if !$d->{node
} || $d->{node
} ne $nodename;
2687 next if !$d->{type
} || $d->{type
} ne 'qemu';
2688 $res->{$vmid}->{exists} = 1;
2693 # test if VM uses local resources (to prevent migration)
2694 sub check_local_resources
{
2695 my ($conf, $noerr) = @_;
2699 push @loc_res, "hostusb" if $conf->{hostusb
}; # old syntax
2700 push @loc_res, "hostpci" if $conf->{hostpci
}; # old syntax
2702 push @loc_res, "ivshmem" if $conf->{ivshmem
};
2704 foreach my $k (keys %$conf) {
2705 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2706 # sockets are safe: they will recreated be on the target side post-migrate
2707 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2708 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2711 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2716 # check if used storages are available on all nodes (use by migrate)
2717 sub check_storage_availability
{
2718 my ($storecfg, $conf, $node) = @_;
2720 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2721 my ($ds, $drive) = @_;
2723 my $volid = $drive->{file
};
2726 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2729 # check if storage is available on both nodes
2730 my $scfg = PVE
::Storage
::storage_check_enabled
($storecfg, $sid);
2731 PVE
::Storage
::storage_check_enabled
($storecfg, $sid, $node);
2733 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $volid);
2735 die "$volid: content type '$vtype' is not available on storage '$sid'\n"
2736 if !$scfg->{content
}->{$vtype};
2740 # list nodes where all VM images are available (used by has_feature API)
2742 my ($conf, $storecfg) = @_;
2744 my $nodelist = PVE
::Cluster
::get_nodelist
();
2745 my $nodehash = { map { $_ => 1 } @$nodelist };
2746 my $nodename = nodename
();
2748 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2749 my ($ds, $drive) = @_;
2751 my $volid = $drive->{file
};
2754 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2756 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2757 if ($scfg->{disable
}) {
2759 } elsif (my $avail = $scfg->{nodes
}) {
2760 foreach my $node (keys %$nodehash) {
2761 delete $nodehash->{$node} if !$avail->{$node};
2763 } elsif (!$scfg->{shared
}) {
2764 foreach my $node (keys %$nodehash) {
2765 delete $nodehash->{$node} if $node ne $nodename
2774 sub check_local_storage_availability
{
2775 my ($conf, $storecfg) = @_;
2777 my $nodelist = PVE
::Cluster
::get_nodelist
();
2778 my $nodehash = { map { $_ => {} } @$nodelist };
2780 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2781 my ($ds, $drive) = @_;
2783 my $volid = $drive->{file
};
2786 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2788 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2790 if ($scfg->{disable
}) {
2791 foreach my $node (keys %$nodehash) {
2792 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2794 } elsif (my $avail = $scfg->{nodes
}) {
2795 foreach my $node (keys %$nodehash) {
2796 if (!$avail->{$node}) {
2797 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2804 foreach my $node (values %$nodehash) {
2805 if (my $unavail = $node->{unavailable_storages
}) {
2806 $node->{unavailable_storages
} = [ sort keys %$unavail ];
2813 # Compat only, use assert_config_exists_on_node and vm_running_locally where possible
2815 my ($vmid, $nocheck, $node) = @_;
2817 # $nocheck is set when called during a migration, in which case the config
2818 # file might still or already reside on the *other* node
2819 # - because rename has already happened, and current node is source
2820 # - because rename hasn't happened yet, and current node is target
2821 # - because rename has happened, current node is target, but hasn't yet
2823 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid, $node) if !$nocheck;
2824 return PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
2829 my $vzlist = config_list
();
2831 my $fd = IO
::Dir-
>new($PVE::QemuServer
::Helpers
::var_run_tmpdir
) || return $vzlist;
2833 while (defined(my $de = $fd->read)) {
2834 next if $de !~ m/^(\d+)\.pid$/;
2836 next if !defined($vzlist->{$vmid});
2837 if (my $pid = check_running
($vmid)) {
2838 $vzlist->{$vmid}->{pid
} = $pid;
2845 our $vmstatus_return_properties = {
2846 vmid
=> get_standard_option
('pve-vmid'),
2848 description
=> "QEMU process status.",
2850 enum
=> ['stopped', 'running'],
2853 description
=> "Maximum memory in bytes.",
2856 renderer
=> 'bytes',
2859 description
=> "Root disk size in bytes.",
2862 renderer
=> 'bytes',
2865 description
=> "VM name.",
2870 description
=> "QEMU QMP agent status.",
2875 description
=> "PID of running qemu process.",
2880 description
=> "Uptime.",
2883 renderer
=> 'duration',
2886 description
=> "Maximum usable CPUs.",
2891 description
=> "The current config lock, if any.",
2896 description
=> "The current configured tags, if any",
2900 'running-machine' => {
2901 description
=> "The currently running machine type (if running).",
2906 description
=> "The currently running QEMU version (if running).",
2912 my $last_proc_pid_stat;
2914 # get VM status information
2915 # This must be fast and should not block ($full == false)
2916 # We only query KVM using QMP if $full == true (this can be slow)
2918 my ($opt_vmid, $full) = @_;
2922 my $storecfg = PVE
::Storage
::config
();
2924 my $list = vzlist
();
2925 my $defaults = load_defaults
();
2927 my ($uptime) = PVE
::ProcFSTools
::read_proc_uptime
(1);
2929 my $cpucount = $cpuinfo->{cpus
} || 1;
2931 foreach my $vmid (keys %$list) {
2932 next if $opt_vmid && ($vmid ne $opt_vmid);
2934 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2936 my $d = { vmid
=> int($vmid) };
2937 $d->{pid
} = int($list->{$vmid}->{pid
}) if $list->{$vmid}->{pid
};
2939 # fixme: better status?
2940 $d->{status
} = $list->{$vmid}->{pid
} ?
'running' : 'stopped';
2942 my $size = PVE
::QemuServer
::Drive
::bootdisk_size
($storecfg, $conf);
2943 if (defined($size)) {
2944 $d->{disk
} = 0; # no info available
2945 $d->{maxdisk
} = $size;
2951 $d->{cpus
} = ($conf->{sockets
} || $defaults->{sockets
})
2952 * ($conf->{cores
} || $defaults->{cores
});
2953 $d->{cpus
} = $cpucount if $d->{cpus
} > $cpucount;
2954 $d->{cpus
} = $conf->{vcpus
} if $conf->{vcpus
};
2956 $d->{name
} = $conf->{name
} || "VM $vmid";
2957 $d->{maxmem
} = $conf->{memory
} ?
$conf->{memory
}*(1024*1024)
2958 : $defaults->{memory
}*(1024*1024);
2960 if ($conf->{balloon
}) {
2961 $d->{balloon_min
} = $conf->{balloon
}*(1024*1024);
2962 $d->{shares
} = defined($conf->{shares
}) ?
$conf->{shares
}
2963 : $defaults->{shares
};
2974 $d->{diskwrite
} = 0;
2976 $d->{template
} = 1 if PVE
::QemuConfig-
>is_template($conf);
2978 $d->{serial
} = 1 if conf_has_serial
($conf);
2979 $d->{lock} = $conf->{lock} if $conf->{lock};
2980 $d->{tags
} = $conf->{tags
} if defined($conf->{tags
});
2985 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
2986 foreach my $dev (keys %$netdev) {
2987 next if $dev !~ m/^tap([1-9]\d*)i/;
2989 my $d = $res->{$vmid};
2992 $d->{netout
} += $netdev->{$dev}->{receive
};
2993 $d->{netin
} += $netdev->{$dev}->{transmit
};
2996 $d->{nics
}->{$dev}->{netout
} = int($netdev->{$dev}->{receive
});
2997 $d->{nics
}->{$dev}->{netin
} = int($netdev->{$dev}->{transmit
});
3002 my $ctime = gettimeofday
;
3004 foreach my $vmid (keys %$list) {
3006 my $d = $res->{$vmid};
3007 my $pid = $d->{pid
};
3010 my $pstat = PVE
::ProcFSTools
::read_proc_pid_stat
($pid);
3011 next if !$pstat; # not running
3013 my $used = $pstat->{utime} + $pstat->{stime
};
3015 $d->{uptime
} = int(($uptime - $pstat->{starttime
})/$cpuinfo->{user_hz
});
3017 if ($pstat->{vsize
}) {
3018 $d->{mem
} = int(($pstat->{rss
}/$pstat->{vsize
})*$d->{maxmem
});
3021 my $old = $last_proc_pid_stat->{$pid};
3023 $last_proc_pid_stat->{$pid} = {
3031 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
3033 if ($dtime > 1000) {
3034 my $dutime = $used - $old->{used
};
3036 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
3037 $last_proc_pid_stat->{$pid} = {
3043 $d->{cpu
} = $old->{cpu
};
3047 return $res if !$full;
3049 my $qmpclient = PVE
::QMPClient-
>new();
3051 my $ballooncb = sub {
3052 my ($vmid, $resp) = @_;
3054 my $info = $resp->{'return'};
3055 return if !$info->{max_mem
};
3057 my $d = $res->{$vmid};
3059 # use memory assigned to VM
3060 $d->{maxmem
} = $info->{max_mem
};
3061 $d->{balloon
} = $info->{actual
};
3063 if (defined($info->{total_mem
}) && defined($info->{free_mem
})) {
3064 $d->{mem
} = $info->{total_mem
} - $info->{free_mem
};
3065 $d->{freemem
} = $info->{free_mem
};
3068 $d->{ballooninfo
} = $info;
3071 my $blockstatscb = sub {
3072 my ($vmid, $resp) = @_;
3073 my $data = $resp->{'return'} || [];
3074 my $totalrdbytes = 0;
3075 my $totalwrbytes = 0;
3077 for my $blockstat (@$data) {
3078 $totalrdbytes = $totalrdbytes + $blockstat->{stats
}->{rd_bytes
};
3079 $totalwrbytes = $totalwrbytes + $blockstat->{stats
}->{wr_bytes
};
3081 $blockstat->{device
} =~ s/drive-//;
3082 $res->{$vmid}->{blockstat
}->{$blockstat->{device
}} = $blockstat->{stats
};
3084 $res->{$vmid}->{diskread
} = $totalrdbytes;
3085 $res->{$vmid}->{diskwrite
} = $totalwrbytes;
3088 my $machinecb = sub {
3089 my ($vmid, $resp) = @_;
3090 my $data = $resp->{'return'} || [];
3092 $res->{$vmid}->{'running-machine'} =
3093 PVE
::QemuServer
::Machine
::current_from_query_machines
($data);
3096 my $versioncb = sub {
3097 my ($vmid, $resp) = @_;
3098 my $data = $resp->{'return'} // {};
3099 my $version = 'unknown';
3101 if (my $v = $data->{qemu
}) {
3102 $version = $v->{major
} . "." . $v->{minor
} . "." . $v->{micro
};
3105 $res->{$vmid}->{'running-qemu'} = $version;
3108 my $statuscb = sub {
3109 my ($vmid, $resp) = @_;
3111 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3112 $qmpclient->queue_cmd($vmid, $machinecb, 'query-machines');
3113 $qmpclient->queue_cmd($vmid, $versioncb, 'query-version');
3114 # this fails if ballon driver is not loaded, so this must be
3115 # the last commnand (following command are aborted if this fails).
3116 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3118 my $status = 'unknown';
3119 if (!defined($status = $resp->{'return'}->{status
})) {
3120 warn "unable to get VM status\n";
3124 $res->{$vmid}->{qmpstatus
} = $resp->{'return'}->{status
};
3127 foreach my $vmid (keys %$list) {
3128 next if $opt_vmid && ($vmid ne $opt_vmid);
3129 next if !$res->{$vmid}->{pid
}; # not running
3130 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3133 $qmpclient->queue_execute(undef, 2);
3135 foreach my $vmid (keys %$list) {
3136 next if $opt_vmid && ($vmid ne $opt_vmid);
3137 next if !$res->{$vmid}->{pid
}; #not running
3139 # we can't use the $qmpclient since it might have already aborted on
3140 # 'query-balloon', but this might also fail for older versions...
3141 my $qemu_support = eval { mon_cmd
($vmid, "query-proxmox-support") };
3142 $res->{$vmid}->{'proxmox-support'} = $qemu_support // {};
3145 foreach my $vmid (keys %$list) {
3146 next if $opt_vmid && ($vmid ne $opt_vmid);
3147 $res->{$vmid}->{qmpstatus
} = $res->{$vmid}->{status
} if !$res->{$vmid}->{qmpstatus
};
3153 sub conf_has_serial
{
3156 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3157 if ($conf->{"serial$i"}) {
3165 sub conf_has_audio
{
3166 my ($conf, $id) = @_;
3169 my $audio = $conf->{"audio$id"};
3170 return if !defined($audio);
3172 my $audioproperties = parse_property_string
($audio_fmt, $audio);
3173 my $audiodriver = $audioproperties->{driver
} // 'spice';
3176 dev
=> $audioproperties->{device
},
3177 dev_id
=> "audiodev$id",
3178 backend
=> $audiodriver,
3179 backend_id
=> "$audiodriver-backend${id}",
3184 my ($audio, $audiopciaddr, $machine_version) = @_;
3188 my $id = $audio->{dev_id
};
3190 if (min_version
($machine_version, 4, 2)) {
3191 $audiodev = ",audiodev=$audio->{backend_id}";
3194 if ($audio->{dev
} eq 'AC97') {
3195 push @$devs, '-device', "AC97,id=${id}${audiopciaddr}$audiodev";
3196 } elsif ($audio->{dev
} =~ /intel\-hda$/) {
3197 push @$devs, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3198 push @$devs, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0$audiodev";
3199 push @$devs, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1$audiodev";
3201 die "unkown audio device '$audio->{dev}', implement me!";
3204 push @$devs, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3212 socket => "/var/run/qemu-server/$vmid.swtpm",
3213 pid
=> "/var/run/qemu-server/$vmid.swtpm.pid",
3217 sub add_tpm_device
{
3218 my ($vmid, $devices, $conf) = @_;
3220 return if !$conf->{tpmstate0
};
3222 my $paths = get_tpm_paths
($vmid);
3224 push @$devices, "-chardev", "socket,id=tpmchar,path=$paths->{socket}";
3225 push @$devices, "-tpmdev", "emulator,id=tpmdev,chardev=tpmchar";
3226 push @$devices, "-device", "tpm-tis,tpmdev=tpmdev";
3230 my ($storecfg, $vmid, $tpmdrive, $migration) = @_;
3232 return if !$tpmdrive;
3235 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
3236 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
3238 $state = PVE
::Storage
::map_volume
($storecfg, $tpm->{file
});
3240 $state = $tpm->{file
};
3243 my $paths = get_tpm_paths
($vmid);
3245 # during migration, we will get state from remote
3248 # run swtpm_setup to create a new TPM state if it doesn't exist yet
3255 "--create-platform-cert",
3258 "/etc/swtpm_setup.conf", # do not use XDG configs
3260 "0", # force creation as root, error if not possible
3261 "--not-overwrite", # ignore existing state, do not modify
3264 push @$setup_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3265 # TPM 2.0 supports ECC crypto, use if possible
3266 push @$setup_cmd, "--ecc" if $tpm->{version
} eq 'v2.0';
3268 run_command
($setup_cmd, outfunc
=> sub {
3269 print "swtpm_setup: $1\n";
3273 # Used to distinguish different invocations in the log.
3274 my $log_prefix = "[id=" . int(time()) . "] ";
3276 my $emulator_cmd = [
3280 "backend-uri=file://$state,mode=0600",
3282 "type=unixio,path=$paths->{socket},mode=0600",
3284 "file=$paths->{pid}",
3285 "--terminate", # terminate on QEMU disconnect
3288 "file=/run/qemu-server/$vmid-swtpm.log,level=1,prefix=$log_prefix",
3290 push @$emulator_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3291 run_command
($emulator_cmd, outfunc
=> sub { print $1; });
3293 my $tries = 100; # swtpm may take a bit to start before daemonizing, wait up to 5s for pid
3294 while (! -e
$paths->{pid
}) {
3295 die "failed to start swtpm: pid file '$paths->{pid}' wasn't created.\n" if --$tries == 0;
3299 # return untainted PID of swtpm daemon so it can be killed on error
3300 file_read_firstline
($paths->{pid
}) =~ m/(\d+)/;
3304 sub vga_conf_has_spice
{
3307 my $vgaconf = parse_vga
($vga);
3308 my $vgatype = $vgaconf->{type
};
3309 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3316 return get_host_arch
() eq $arch;
3321 return $conf->{arch
} // get_host_arch
();
3324 my $default_machines = {
3329 sub get_installed_machine_version
{
3330 my ($kvmversion) = @_;
3331 $kvmversion = kvm_user_version
() if !defined($kvmversion);
3332 $kvmversion =~ m/^(\d+\.\d+)/;
3336 sub windows_get_pinned_machine_version
{
3337 my ($machine, $base_version, $kvmversion) = @_;
3339 my $pin_version = $base_version;
3340 if (!defined($base_version) ||
3341 !PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($base_version, $kvmversion)
3343 $pin_version = get_installed_machine_version
($kvmversion);
3345 if (!$machine || $machine eq 'pc') {
3346 $machine = "pc-i440fx-$pin_version";
3347 } elsif ($machine eq 'q35') {
3348 $machine = "pc-q35-$pin_version";
3349 } elsif ($machine eq 'virt') {
3350 $machine = "virt-$pin_version";
3352 warn "unknown machine type '$machine', not touching that!\n";
3358 sub get_vm_machine
{
3359 my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
3361 my $machine = $forcemachine || $conf->{machine
};
3363 if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
3364 $kvmversion //= kvm_user_version
();
3365 # we must pin Windows VMs without a specific version to 5.1, as 5.2 fixed a bug in ACPI
3366 # layout which confuses windows quite a bit and may result in various regressions..
3367 # see: https://lists.gnu.org/archive/html/qemu-devel/2021-02/msg08484.html
3368 if (windows_version
($conf->{ostype
})) {
3369 $machine = windows_get_pinned_machine_version
($machine, '5.1', $kvmversion);
3372 $machine ||= $default_machines->{$arch};
3373 if ($add_pve_version) {
3374 my $pvever = PVE
::QemuServer
::Machine
::get_pve_version
($kvmversion);
3375 $machine .= "+pve$pvever";
3379 if ($add_pve_version && $machine !~ m/\+pve\d+?(?:\.pxe)?$/) {
3380 my $is_pxe = $machine =~ m/^(.*?)\.pxe$/;
3381 $machine = $1 if $is_pxe;
3383 # for version-pinned machines that do not include a pve-version (e.g.
3384 # pc-q35-4.1), we assume 0 to keep them stable in case we bump
3385 $machine .= '+pve0';
3387 $machine .= '.pxe' if $is_pxe;
3393 sub get_ovmf_files
($$$) {
3394 my ($arch, $efidisk, $smm) = @_;
3396 my $types = $OVMF->{$arch}
3397 or die "no OVMF images known for architecture '$arch'\n";
3399 my $type = 'default';
3400 if ($arch ne "aarch64" && defined($efidisk->{efitype
}) && $efidisk->{efitype
} eq '4m') {
3401 $type = $smm ?
"4m" : "4m-no-smm";
3402 $type .= '-ms' if $efidisk->{'pre-enrolled-keys'};
3405 my ($ovmf_code, $ovmf_vars) = $types->{$type}->@*;
3406 die "EFI base image '$ovmf_code' not found\n" if ! -f
$ovmf_code;
3407 die "EFI vars image '$ovmf_vars' not found\n" if ! -f
$ovmf_vars;
3409 return ($ovmf_code, $ovmf_vars);
3413 aarch64
=> '/usr/bin/qemu-system-aarch64',
3414 x86_64
=> '/usr/bin/qemu-system-x86_64',
3416 sub get_command_for_arch
($) {
3418 return '/usr/bin/kvm' if is_native
($arch);
3420 my $cmd = $Arch2Qemu->{$arch}
3421 or die "don't know how to emulate architecture '$arch'\n";
3425 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
3426 # to use in a QEMU command line (-cpu element), first array_intersect the result
3427 # of query_supported_ with query_understood_. This is necessary because:
3429 # a) query_understood_ returns flags the host cannot use and
3430 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
3431 # flags, but CPU settings - with most of them being flags. Those settings
3432 # (and some flags, curiously) cannot be specified as a "-cpu" argument.
3434 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
3435 # expensive. If you need the value returned from this, you can get it much
3436 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
3437 # $accel being 'kvm' or 'tcg'.
3439 # pvestatd calls this function on startup and whenever the QEMU/KVM version
3440 # changes, automatically populating pmxcfs.
3442 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
3443 # since kvm and tcg machines support different flags
3445 sub query_supported_cpu_flags
{
3448 $arch //= get_host_arch
();
3449 my $default_machine = $default_machines->{$arch};
3453 # FIXME: Once this is merged, the code below should work for ARM as well:
3454 # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
3455 die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
3458 my $kvm_supported = defined(kvm_version
());
3459 my $qemu_cmd = get_command_for_arch
($arch);
3461 my $pidfile = PVE
::QemuServer
::Helpers
::pidfile_name
($fakevmid);
3463 # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
3464 my $query_supported_run_qemu = sub {
3470 '-machine', $default_machine,
3472 '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server=on,wait=off",
3473 '-mon', 'chardev=qmp,mode=control',
3474 '-pidfile', $pidfile,
3479 push @$cmd, '-accel', 'tcg';
3482 my $rc = run_command
($cmd, noerr
=> 1, quiet
=> 0);
3483 die "QEMU flag querying VM exited with code " . $rc if $rc;
3486 my $cmd_result = mon_cmd
(
3488 'query-cpu-model-expansion',
3490 model
=> { name
=> 'host' }
3493 my $props = $cmd_result->{model
}->{props
};
3494 foreach my $prop (keys %$props) {
3495 next if $props->{$prop} ne '1';
3496 # QEMU returns some flags multiple times, with '_', '.' or '-'
3497 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
3498 # We only keep those with underscores, to match /proc/cpuinfo
3499 $prop =~ s/\.|-/_/g;
3500 $flags->{$prop} = 1;
3505 # force stop with 10 sec timeout and 'nocheck', always stop, even if QMP failed
3506 vm_stop
(undef, $fakevmid, 1, 1, 10, 0, 1);
3510 return [ sort keys %$flags ];
3513 # We need to query QEMU twice, since KVM and TCG have different supported flags
3514 PVE
::QemuConfig-
>lock_config($fakevmid, sub {
3515 $flags->{tcg
} = eval { $query_supported_run_qemu->(0) };
3516 warn "warning: failed querying supported tcg flags: $@\n" if $@;
3518 if ($kvm_supported) {
3519 $flags->{kvm
} = eval { $query_supported_run_qemu->(1) };
3520 warn "warning: failed querying supported kvm flags: $@\n" if $@;
3527 # Understood CPU flags are written to a file at 'pve-qemu' compile time
3528 my $understood_cpu_flag_dir = "/usr/share/kvm";
3529 sub query_understood_cpu_flags
{
3530 my $arch = get_host_arch
();
3531 my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
3533 die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
3536 my $raw = file_get_contents
($filepath);
3537 $raw =~ s/^\s+|\s+$//g;
3538 my @flags = split(/\s+/, $raw);
3543 # Since commit 277d33454f77ec1d1e0bc04e37621e4dd2424b67 in pve-qemu, smm is not off by default
3544 # anymore. But smm=off seems to be required when using SeaBIOS and serial display.
3545 my sub should_disable_smm
{
3546 my ($conf, $vga) = @_;
3548 return (!defined($conf->{bios
}) || $conf->{bios
} eq 'seabios') &&
3549 $vga->{type
} && $vga->{type
} =~ m/^(serial\d+|none)$/;
3552 my sub print_ovmf_drive_commandlines
{
3553 my ($conf, $storecfg, $vmid, $arch, $q35, $version_guard) = @_;
3555 my $d = $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
3557 my ($ovmf_code, $ovmf_vars) = get_ovmf_files
($arch, $d, $q35);
3559 my $var_drive_str = "if=pflash,unit=1,id=drive-efidisk0";
3561 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($d->{file
}, 1);
3562 my ($path, $format) = $d->@{'file', 'format'};
3564 $path = PVE
::Storage
::path
($storecfg, $d->{file
});
3565 if (!defined($format)) {
3566 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
3567 $format = qemu_img_format
($scfg, $volname);
3569 } elsif (!defined($format)) {
3570 die "efidisk format must be specified\n";
3572 # SPI flash does lots of read-modify-write OPs, without writeback this gets really slow #3329
3573 if ($path =~ m/^rbd:/) {
3574 $var_drive_str .= ',cache=writeback';
3575 $path .= ':rbd_cache_policy=writeback'; # avoid write-around, we *need* to cache writes too
3577 $var_drive_str .= ",format=$format,file=$path";
3579 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $format eq 'raw' && $version_guard->(4, 1, 2);
3580 $var_drive_str .= ',readonly=on' if drive_is_read_only
($conf, $d);
3582 log_warn
("no efidisk configured! Using temporary efivars disk.");
3583 my $path = "/tmp/$vmid-ovmf.fd";
3584 PVE
::Tools
::file_copy
($ovmf_vars, $path, -s
$ovmf_vars);
3585 $var_drive_str .= ",format=raw,file=$path";
3586 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $version_guard->(4, 1, 2);
3589 return ("if=pflash,unit=0,format=raw,readonly=on,file=$ovmf_code", $var_drive_str);
3592 sub config_to_command
{
3593 my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu,
3596 my ($globalFlags, $machineFlags, $rtcFlags) = ([], [], []);
3599 my $ostype = $conf->{ostype
};
3600 my $winversion = windows_version
($ostype);
3601 my $kvm = $conf->{kvm
};
3602 my $nodename = nodename
();
3604 my $arch = get_vm_arch
($conf);
3605 my $kvm_binary = get_command_for_arch
($arch);
3606 my $kvmver = kvm_user_version
($kvm_binary);
3608 if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
3609 $kvmver //= "undefined";
3610 die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
3613 my $add_pve_version = min_version
($kvmver, 4, 1);
3615 my $machine_type = get_vm_machine
($conf, $forcemachine, $arch, $add_pve_version);
3616 my $machine_version = extract_version
($machine_type, $kvmver);
3617 $kvm //= 1 if is_native
($arch);
3619 $machine_version =~ m/(\d+)\.(\d+)/;
3620 my ($machine_major, $machine_minor) = ($1, $2);
3622 if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
3623 warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
3624 } elsif (!min_version
($kvmver, $machine_major, $machine_minor)) {
3625 die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type',"
3626 ." please upgrade node '$nodename'\n"
3627 } elsif (!PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($machine_version, $kvmver)) {
3628 my $max_pve_version = PVE
::QemuServer
::Machine
::get_pve_version
($machine_version);
3629 die "Installed qemu-server (max feature level for $machine_major.$machine_minor is"
3630 ." pve$max_pve_version) is too old to run machine type '$machine_type', please upgrade"
3631 ." node '$nodename'\n";
3634 # if a specific +pve version is required for a feature, use $version_guard
3635 # instead of min_version to allow machines to be run with the minimum
3637 my $required_pve_version = 0;
3638 my $version_guard = sub {
3639 my ($major, $minor, $pve) = @_;
3640 return 0 if !min_version
($machine_version, $major, $minor, $pve);
3641 my $max_pve = PVE
::QemuServer
::Machine
::get_pve_version
("$major.$minor");
3642 return 1 if min_version
($machine_version, $major, $minor, $max_pve+1);
3643 $required_pve_version = $pve if $pve && $pve > $required_pve_version;
3647 if ($kvm && !defined kvm_version
()) {
3648 die "KVM virtualisation configured, but not available. Either disable in VM configuration"
3649 ." or enable in BIOS.\n";
3652 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
3653 my $hotplug_features = parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
3654 my $use_old_bios_files = undef;
3655 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
3658 if ($conf->{affinity
}) {
3659 push @$cmd, '/usr/bin/taskset', '--cpu-list', '--all-tasks', $conf->{affinity
};
3662 push @$cmd, $kvm_binary;
3664 push @$cmd, '-id', $vmid;
3666 my $vmname = $conf->{name
} || "vm$vmid";
3668 push @$cmd, '-name', "$vmname,debug-threads=on";
3670 push @$cmd, '-no-shutdown';
3674 my $qmpsocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid);
3675 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server=on,wait=off";
3676 push @$cmd, '-mon', "chardev=qmp,mode=control";
3678 if (min_version
($machine_version, 2, 12)) {
3679 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3680 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3683 push @$cmd, '-pidfile' , PVE
::QemuServer
::Helpers
::pidfile_name
($vmid);
3685 push @$cmd, '-daemonize';
3687 if ($conf->{smbios1
}) {
3688 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
3689 if ($smbios_conf->{base64
}) {
3690 # Do not pass base64 flag to qemu
3691 delete $smbios_conf->{base64
};
3692 my $smbios_string = "";
3693 foreach my $key (keys %$smbios_conf) {
3695 if ($key eq "uuid") {
3696 $value = $smbios_conf->{uuid
}
3698 $value = decode_base64
($smbios_conf->{$key});
3700 # qemu accepts any binary data, only commas need escaping by double comma
3702 $smbios_string .= "," . $key . "=" . $value if $value;
3704 push @$cmd, '-smbios', "type=1" . $smbios_string;
3706 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3710 if ($conf->{bios
} && $conf->{bios
} eq 'ovmf') {
3711 my ($code_drive_str, $var_drive_str) =
3712 print_ovmf_drive_commandlines
($conf, $storecfg, $vmid, $arch, $q35, $version_guard);
3713 push $cmd->@*, '-drive', $code_drive_str;
3714 push $cmd->@*, '-drive', $var_drive_str;
3717 if ($q35) { # tell QEMU to load q35 config early
3718 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3719 if (min_version
($machine_version, 4, 0)) {
3720 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3722 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3726 if (defined(my $fixups = qemu_created_version_fixups
($conf, $forcemachine, $kvmver))) {
3727 push @$cmd, $fixups->@*;
3730 if ($conf->{vmgenid
}) {
3731 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid
};
3734 # add usb controllers
3735 my @usbcontrollers = PVE
::QemuServer
::USB
::get_usb_controllers
(
3736 $conf, $bridges, $arch, $machine_type, $usbdesc->{format
}, $MAX_USB_DEVICES, $machine_version);
3737 push @$devices, @usbcontrollers if @usbcontrollers;
3738 my $vga = parse_vga
($conf->{vga
});
3740 my $qxlnum = vga_conf_has_spice
($conf->{vga
});
3741 $vga->{type
} = 'qxl' if $qxlnum;
3743 if (!$vga->{type
}) {
3744 if ($arch eq 'aarch64') {
3745 $vga->{type
} = 'virtio';
3746 } elsif (min_version
($machine_version, 2, 9)) {
3747 $vga->{type
} = (!$winversion || $winversion >= 6) ?
'std' : 'cirrus';
3749 $vga->{type
} = ($winversion >= 6) ?
'std' : 'cirrus';
3753 # enable absolute mouse coordinates (needed by vnc)
3754 my $tablet = $conf->{tablet
};
3755 if (!defined($tablet)) {
3756 $tablet = $defaults->{tablet
};
3757 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3758 $tablet = 0 if $vga->{type
} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3762 push @$devices, '-device', print_tabletdevice_full
($conf, $arch) if $tablet;
3763 my $kbd = print_keyboarddevice_full
($conf, $arch);
3764 push @$devices, '-device', $kbd if defined($kbd);
3767 my $bootorder = device_bootorder
($conf);
3769 # host pci device passthrough
3770 my ($kvm_off, $gpu_passthrough, $legacy_igd) = PVE
::QemuServer
::PCI
::print_hostpci_devices
(
3771 $vmid, $conf, $devices, $vga, $winversion, $q35, $bridges, $arch, $machine_type, $bootorder);
3774 my $usb_dev_features = {};
3775 $usb_dev_features->{spice_usb3
} = 1 if min_version
($machine_version, 4, 0);
3777 my @usbdevices = PVE
::QemuServer
::USB
::get_usb_devices
(
3778 $conf, $usbdesc->{format
}, $MAX_USB_DEVICES, $usb_dev_features, $bootorder, $machine_version);
3779 push @$devices, @usbdevices if @usbdevices;
3782 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3783 my $path = $conf->{"serial$i"} or next;
3784 if ($path eq 'socket') {
3785 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3786 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server=on,wait=off";
3787 # On aarch64, serial0 is the UART device. QEMU only allows
3788 # connecting UART devices via the '-serial' command line, as
3789 # the device has a fixed slot on the hardware...
3790 if ($arch eq 'aarch64' && $i == 0) {
3791 push @$devices, '-serial', "chardev:serial$i";
3793 push @$devices, '-device', "isa-serial,chardev=serial$i";
3796 die "no such serial device\n" if ! -c
$path;
3797 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3798 push @$devices, '-device', "isa-serial,chardev=serial$i";
3803 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3804 if (my $path = $conf->{"parallel$i"}) {
3805 die "no such parallel device\n" if ! -c
$path;
3806 my $devtype = $path =~ m!^/dev/usb/lp! ?
'tty' : 'parport';
3807 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3808 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3812 if (min_version
($machine_version, 4, 0) && (my $audio = conf_has_audio
($conf))) {
3813 my $audiopciaddr = print_pci_addr
("audio0", $bridges, $arch, $machine_type);
3814 my $audio_devs = audio_devs
($audio, $audiopciaddr, $machine_version);
3815 push @$devices, @$audio_devs;
3818 add_tpm_device
($vmid, $devices, $conf);
3821 $sockets = $conf->{smp
} if $conf->{smp
}; # old style - no longer iused
3822 $sockets = $conf->{sockets
} if $conf->{sockets
};
3824 my $cores = $conf->{cores
} || 1;
3826 my $maxcpus = $sockets * $cores;
3828 my $vcpus = $conf->{vcpus
} ?
$conf->{vcpus
} : $maxcpus;
3830 my $allowed_vcpus = $cpuinfo->{cpus
};
3832 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n" if ($allowed_vcpus < $maxcpus);
3834 if ($hotplug_features->{cpu
} && min_version
($machine_version, 2, 7)) {
3835 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3836 for (my $i = 2; $i <= $vcpus; $i++) {
3837 my $cpustr = print_cpu_device
($conf,$i);
3838 push @$cmd, '-device', $cpustr;
3843 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3845 push @$cmd, '-nodefaults';
3847 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3849 push @$cmd, '-no-acpi' if defined($conf->{acpi
}) && $conf->{acpi
} == 0;
3851 push @$cmd, '-no-reboot' if defined($conf->{reboot
}) && $conf->{reboot
} == 0;
3853 if ($vga->{type
} && $vga->{type
} !~ m/^serial\d+$/ && $vga->{type
} ne 'none'){
3854 push @$devices, '-device', print_vga_device
(
3855 $conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3857 push @$cmd, '-display', 'egl-headless,gl=core' if $vga->{type
} eq 'virtio-gl'; # VIRGL
3859 my $socket = PVE
::QemuServer
::Helpers
::vnc_socket
($vmid);
3860 push @$cmd, '-vnc', "unix:$socket,password=on";
3862 push @$cmd, '-vga', 'none' if $vga->{type
} eq 'none';
3863 push @$cmd, '-nographic';
3867 my $tdf = defined($conf->{tdf
}) ?
$conf->{tdf
} : $defaults->{tdf
};
3868 my $useLocaltime = $conf->{localtime};
3870 if ($winversion >= 5) { # windows
3871 $useLocaltime = 1 if !defined($conf->{localtime});
3873 # use time drift fix when acpi is enabled
3874 if (!(defined($conf->{acpi
}) && $conf->{acpi
} == 0)) {
3875 $tdf = 1 if !defined($conf->{tdf
});
3879 if ($winversion >= 6) {
3880 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3881 push @$cmd, '-no-hpet';
3884 push @$rtcFlags, 'driftfix=slew' if $tdf;
3886 if ($conf->{startdate
} && $conf->{startdate
} ne 'now') {
3887 push @$rtcFlags, "base=$conf->{startdate}";
3888 } elsif ($useLocaltime) {
3889 push @$rtcFlags, 'base=localtime';
3893 push @$cmd, '-cpu', $forcecpu;
3895 push @$cmd, get_cpu_options
($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough);
3898 PVE
::QemuServer
::Memory
::config
(
3899 $conf, $vmid, $sockets, $cores, $defaults, $hotplug_features->{memory
}, $cmd);
3901 push @$cmd, '-S' if $conf->{freeze
};
3903 push @$cmd, '-k', $conf->{keyboard
} if defined($conf->{keyboard
});
3905 my $guest_agent = parse_guest_agent
($conf);
3907 if ($guest_agent->{enabled
}) {
3908 my $qgasocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid, 1);
3909 push @$devices, '-chardev', "socket,path=$qgasocket,server=on,wait=off,id=qga0";
3911 if (!$guest_agent->{type
} || $guest_agent->{type
} eq 'virtio') {
3912 my $pciaddr = print_pci_addr
("qga0", $bridges, $arch, $machine_type);
3913 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3914 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3915 } elsif ($guest_agent->{type
} eq 'isa') {
3916 push @$devices, '-device', "isa-serial,chardev=qga0";
3920 my $rng = $conf->{rng0
} ? parse_rng
($conf->{rng0
}) : undef;
3921 if ($rng && $version_guard->(4, 1, 2)) {
3922 check_rng_source
($rng->{source
});
3924 my $max_bytes = $rng->{max_bytes
} // $rng_fmt->{max_bytes
}->{default};
3925 my $period = $rng->{period
} // $rng_fmt->{period
}->{default};
3926 my $limiter_str = "";
3928 $limiter_str = ",max-bytes=$max_bytes,period=$period";
3931 my $rng_addr = print_pci_addr
("rng0", $bridges, $arch, $machine_type);
3932 push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3933 push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3938 if ($qxlnum || $vga->{type
} =~ /^virtio/) {
3941 for (my $i = 1; $i < $qxlnum; $i++){
3942 push @$devices, '-device', print_vga_device
(
3943 $conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3946 # assume other OS works like Linux
3947 my ($ram, $vram) = ("134217728", "67108864");
3948 if ($vga->{memory
}) {
3949 $ram = PVE
::Tools
::convert_size
($qxlnum*4*$vga->{memory
}, 'mb' => 'b');
3950 $vram = PVE
::Tools
::convert_size
($qxlnum*2*$vga->{memory
}, 'mb' => 'b');
3952 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3953 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3957 my $pciaddr = print_pci_addr
("spice", $bridges, $arch, $machine_type);
3959 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
3960 my @nodeaddrs = PVE
::Tools
::getaddrinfo_all
('localhost', family
=> $pfamily);
3961 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3963 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3964 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3965 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3967 my $localhost = PVE
::Network
::addr_to_ip
($nodeaddrs[0]->{addr
});
3968 $spice_port = PVE
::Tools
::next_spice_port
($pfamily, $localhost);
3970 my $spice_enhancement_str = $conf->{spice_enhancements
} // '';
3971 my $spice_enhancement = parse_property_string
($spice_enhancements_fmt, $spice_enhancement_str);
3972 if ($spice_enhancement->{foldersharing
}) {
3973 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3974 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3977 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3978 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}"
3979 if $spice_enhancement->{videostreaming
};
3981 push @$devices, '-spice', "$spice_opts";
3984 # enable balloon by default, unless explicitly disabled
3985 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
3986 my $pciaddr = print_pci_addr
("balloon0", $bridges, $arch, $machine_type);
3987 my $ballooncmd = "virtio-balloon-pci,id=balloon0$pciaddr";
3988 $ballooncmd .= ",free-page-reporting=on" if min_version
($machine_version, 6, 2);
3989 push @$devices, '-device', $ballooncmd;
3992 if ($conf->{watchdog
}) {
3993 my $wdopts = parse_watchdog
($conf->{watchdog
});
3994 my $pciaddr = print_pci_addr
("watchdog", $bridges, $arch, $machine_type);
3995 my $watchdog = $wdopts->{model
} || 'i6300esb';
3996 push @$devices, '-device', "$watchdog$pciaddr";
3997 push @$devices, '-watchdog-action', $wdopts->{action
} if $wdopts->{action
};
4001 my $scsicontroller = {};
4002 my $ahcicontroller = {};
4003 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : $defaults->{scsihw
};
4005 # Add iscsi initiator name if available
4006 if (my $initiator = get_initiator_name
()) {
4007 push @$devices, '-iscsi', "initiator-name=$initiator";
4010 PVE
::QemuConfig-
>foreach_volume($conf, sub {
4011 my ($ds, $drive) = @_;
4013 if (PVE
::Storage
::parse_volume_id
($drive->{file
}, 1)) {
4014 check_volume_storage_type
($storecfg, $drive->{file
});
4015 push @$vollist, $drive->{file
};
4018 # ignore efidisk here, already added in bios/fw handling code above
4019 return if $drive->{interface
} eq 'efidisk';
4021 return if $drive->{interface
} eq 'tpmstate';
4023 $use_virtio = 1 if $ds =~ m/^virtio/;
4025 $drive->{bootindex
} = $bootorder->{$ds} if $bootorder->{$ds};
4027 if ($drive->{interface
} eq 'virtio'){
4028 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread
};
4031 if ($drive->{interface
} eq 'scsi') {
4033 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
4035 die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
4036 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
4038 my $pciaddr = print_pci_addr
("$controller_prefix$controller", $bridges, $arch, $machine_type);
4039 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ?
"virtio-scsi-pci" : $scsihw;
4042 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{iothread
}){
4043 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4044 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4045 } elsif ($drive->{iothread
}) {
4047 "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n"
4052 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{queues
}){
4053 $queues = ",num_queues=$drive->{queues}";
4056 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues"
4057 if !$scsicontroller->{$controller};
4058 $scsicontroller->{$controller}=1;
4061 if ($drive->{interface
} eq 'sata') {
4062 my $controller = int($drive->{index} / $PVE::QemuServer
::Drive
::MAX_SATA_DISKS
);
4063 my $pciaddr = print_pci_addr
("ahci$controller", $bridges, $arch, $machine_type);
4064 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr"
4065 if !$ahcicontroller->{$controller};
4066 $ahcicontroller->{$controller}=1;
4069 my $pbs_conf = $pbs_backing->{$ds};
4070 my $pbs_name = undef;
4072 $pbs_name = "drive-$ds-pbs";
4073 push @$devices, '-blockdev', print_pbs_blockdev
($pbs_conf, $pbs_name);
4076 my $drive_cmd = print_drive_commandline_full
(
4077 $storecfg, $vmid, $drive, $pbs_name, min_version
($kvmver, 6, 0));
4079 # extra protection for templates, but SATA and IDE don't support it..
4080 $drive_cmd .= ',readonly=on' if drive_is_read_only
($conf, $drive);
4082 push @$devices, '-drive',$drive_cmd;
4083 push @$devices, '-device', print_drivedevice_full
(
4084 $storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4087 for (my $i = 0; $i < $MAX_NETS; $i++) {
4088 my $netname = "net$i";
4090 next if !$conf->{$netname};
4091 my $d = parse_net
($conf->{$netname});
4093 # save the MAC addr here (could be auto-gen. in some odd setups) for FDB registering later?
4095 $use_virtio = 1 if $d->{model
} eq 'virtio';
4097 $d->{bootindex
} = $bootorder->{$netname} if $bootorder->{$netname};
4099 my $netdevfull = print_netdev_full
($vmid, $conf, $arch, $d, $netname);
4100 push @$devices, '-netdev', $netdevfull;
4102 my $netdevicefull = print_netdevice_full
(
4103 $vmid, $conf, $d, $netname, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version);
4105 push @$devices, '-device', $netdevicefull;
4108 if ($conf->{ivshmem
}) {
4109 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
4113 $bus = print_pcie_addr
("ivshmem");
4115 $bus = print_pci_addr
("ivshmem", $bridges, $arch, $machine_type);
4118 my $ivshmem_name = $ivshmem->{name
} // $vmid;
4119 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4121 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4122 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path"
4123 .",size=$ivshmem->{size}M";
4126 # pci.4 is nested in pci.1
4127 $bridges->{1} = 1 if $bridges->{4};
4129 if (!$q35) { # add pci bridges
4130 if (min_version
($machine_version, 2, 3)) {
4134 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4137 for my $k (sort {$b cmp $a} keys %$bridges) {
4138 next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
4141 if ($k == 2 && $legacy_igd) {
4144 my $pciaddr = print_pci_addr
("pci.$k_name", undef, $arch, $machine_type);
4145 my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
4147 if ($q35) { # add after -readconfig pve-q35.cfg
4148 splice @$devices, 2, 0, '-device', $devstr;
4150 unshift @$devices, '-device', $devstr if $k > 0;
4155 push @$machineFlags, 'accel=tcg';
4158 push @$machineFlags, 'smm=off' if should_disable_smm
($conf, $vga);
4160 my $machine_type_min = $machine_type;
4161 if ($add_pve_version) {
4162 $machine_type_min =~ s/\+pve\d+$//;
4163 $machine_type_min .= "+pve$required_pve_version";
4165 push @$machineFlags, "type=${machine_type_min}";
4167 push @$cmd, @$devices;
4168 push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
4169 push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
4170 push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
4172 if (my $vmstate = $conf->{vmstate
}) {
4173 my $statepath = PVE
::Storage
::path
($storecfg, $vmstate);
4174 push @$vollist, $vmstate;
4175 push @$cmd, '-loadstate', $statepath;
4176 print "activating and using '$vmstate' as vmstate\n";
4179 if (PVE
::QemuConfig-
>is_template($conf)) {
4180 # needed to workaround base volumes being read-only
4181 push @$cmd, '-snapshot';
4185 if ($conf->{args
}) {
4186 my $aa = PVE
::Tools
::split_args
($conf->{args
});
4190 return wantarray ?
($cmd, $vollist, $spice_port) : $cmd;
4193 sub check_rng_source
{
4196 # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
4197 die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
4200 my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
4201 if ($source eq '/dev/hwrng' && file_read_firstline
($rng_current) eq 'none') {
4202 # Needs to abort, otherwise QEMU crashes on first rng access. Note that rng_current cannot
4203 # be changed to 'none' manually, so once the VM is past this point, it's no longer an issue.
4204 die "Cannot start VM with passed-through RNG device: '/dev/hwrng' exists, but"
4205 ." '$rng_current' is set to 'none'. Ensure that a compatible hardware-RNG is attached"
4213 my $res = mon_cmd
($vmid, 'query-spice');
4215 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4218 sub vm_devices_list
{
4221 my $res = mon_cmd
($vmid, 'query-pci');
4222 my $devices_to_check = [];
4224 foreach my $pcibus (@$res) {
4225 push @$devices_to_check, @{$pcibus->{devices
}},
4228 while (@$devices_to_check) {
4230 for my $d (@$devices_to_check) {
4231 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4232 next if !$d->{'pci_bridge'} || !$d->{'pci_bridge'}->{devices
};
4234 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices
}});
4235 push @$to_check, @{$d->{'pci_bridge'}->{devices
}};
4237 $devices_to_check = $to_check;
4240 my $resblock = mon_cmd
($vmid, 'query-block');
4241 foreach my $block (@$resblock) {
4242 if($block->{device
} =~ m/^drive-(\S+)/){
4247 my $resmice = mon_cmd
($vmid, 'query-mice');
4248 foreach my $mice (@$resmice) {
4249 if ($mice->{name
} eq 'QEMU HID Tablet') {
4250 $devices->{tablet
} = 1;
4255 # for usb devices there is no query-usb
4256 # but we can iterate over the entries in
4257 # qom-list path=/machine/peripheral
4258 my $resperipheral = mon_cmd
($vmid, 'qom-list', path
=> '/machine/peripheral');
4259 foreach my $per (@$resperipheral) {
4260 if ($per->{name
} =~ m/^usb(?:redirdev)?\d+$/) {
4261 $devices->{$per->{name
}} = 1;
4269 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4271 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
4273 my $devices_list = vm_devices_list
($vmid);
4274 return 1 if defined($devices_list->{$deviceid});
4276 # add PCI bridge if we need it for the device
4277 qemu_add_pci_bridge
($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type);
4279 if ($deviceid eq 'tablet') {
4280 qemu_deviceadd
($vmid, print_tabletdevice_full
($conf, $arch));
4281 } elsif ($deviceid eq 'keyboard') {
4282 qemu_deviceadd
($vmid, print_keyboarddevice_full
($conf, $arch));
4283 } elsif ($deviceid =~ m/^usbredirdev(\d+)$/) {
4285 qemu_spice_usbredir_chardev_add
($vmid, "usbredirchardev$id");
4286 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_spice_usbdevice
($id, "xhci", $id + 1));
4287 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4288 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_usbdevice_full
($conf, $deviceid, $device, {}, $1 + 1));
4289 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4290 qemu_iothread_add
($vmid, $deviceid, $device);
4292 qemu_driveadd
($storecfg, $vmid, $device);
4293 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4295 qemu_deviceadd
($vmid, $devicefull);
4296 eval { qemu_deviceaddverify
($vmid, $deviceid); };
4298 eval { qemu_drivedel
($vmid, $deviceid); };
4302 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4303 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : "lsi";
4304 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4305 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ?
"virtio-scsi-pci" : $scsihw;
4307 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4309 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread
}) {
4310 qemu_iothread_add
($vmid, $deviceid, $device);
4311 $devicefull .= ",iothread=iothread-$deviceid";
4314 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues
}) {
4315 $devicefull .= ",num_queues=$device->{queues}";
4318 qemu_deviceadd
($vmid, $devicefull);
4319 qemu_deviceaddverify
($vmid, $deviceid);
4320 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4321 qemu_findorcreatescsihw
($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4322 qemu_driveadd
($storecfg, $vmid, $device);
4324 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4325 eval { qemu_deviceadd
($vmid, $devicefull); };
4327 eval { qemu_drivedel
($vmid, $deviceid); };
4331 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4332 return if !qemu_netdevadd
($vmid, $conf, $arch, $device, $deviceid);
4334 my $machine_type = PVE
::QemuServer
::Machine
::qemu_machine_pxe
($vmid, $conf);
4335 my $machine_version = PVE
::QemuServer
::Machine
::extract_version
($machine_type);
4336 my $use_old_bios_files = undef;
4337 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
4339 my $netdevicefull = print_netdevice_full
(
4340 $vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type, $machine_version);
4341 qemu_deviceadd
($vmid, $netdevicefull);
4343 qemu_deviceaddverify
($vmid, $deviceid);
4344 qemu_set_link_status
($vmid, $deviceid, !$device->{link_down
});
4347 eval { qemu_netdevdel
($vmid, $deviceid); };
4351 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4353 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4354 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4356 qemu_deviceadd
($vmid, $devicefull);
4357 qemu_deviceaddverify
($vmid, $deviceid);
4359 die "can't hotplug device '$deviceid'\n";
4365 # fixme: this should raise exceptions on error!
4366 sub vm_deviceunplug
{
4367 my ($vmid, $conf, $deviceid) = @_;
4369 my $devices_list = vm_devices_list
($vmid);
4370 return 1 if !defined($devices_list->{$deviceid});
4372 my $bootdisks = PVE
::QemuServer
::Drive
::get_bootdisks
($conf);
4373 die "can't unplug bootdisk '$deviceid'\n" if grep {$_ eq $deviceid} @$bootdisks;
4375 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard' || $deviceid eq 'xhci') {
4376 qemu_devicedel
($vmid, $deviceid);
4377 } elsif ($deviceid =~ m/^usbredirdev\d+$/) {
4378 qemu_devicedel
($vmid, $deviceid);
4379 qemu_devicedelverify
($vmid, $deviceid);
4380 } elsif ($deviceid =~ m/^usb\d+$/) {
4381 qemu_devicedel
($vmid, $deviceid);
4382 qemu_devicedelverify
($vmid, $deviceid);
4383 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4384 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4386 qemu_devicedel
($vmid, $deviceid);
4387 qemu_devicedelverify
($vmid, $deviceid);
4388 qemu_drivedel
($vmid, $deviceid);
4389 qemu_iothread_del
($vmid, $deviceid, $device);
4390 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4391 qemu_devicedel
($vmid, $deviceid);
4392 qemu_devicedelverify
($vmid, $deviceid);
4393 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4394 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4396 qemu_devicedel
($vmid, $deviceid);
4397 qemu_devicedelverify
($vmid, $deviceid);
4398 qemu_drivedel
($vmid, $deviceid);
4399 qemu_deletescsihw
($conf, $vmid, $deviceid);
4401 qemu_iothread_del
($vmid, "virtioscsi$device->{index}", $device)
4402 if $conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single');
4403 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4404 qemu_devicedel
($vmid, $deviceid);
4405 qemu_devicedelverify
($vmid, $deviceid);
4406 qemu_netdevdel
($vmid, $deviceid);
4408 die "can't unplug device '$deviceid'\n";
4414 sub qemu_spice_usbredir_chardev_add
{
4415 my ($vmid, $id) = @_;
4417 mon_cmd
($vmid, "chardev-add" , (
4428 sub qemu_deviceadd
{
4429 my ($vmid, $devicefull) = @_;
4431 $devicefull = "driver=".$devicefull;
4432 my %options = split(/[=,]/, $devicefull);
4434 mon_cmd
($vmid, "device_add" , %options);
4437 sub qemu_devicedel
{
4438 my ($vmid, $deviceid) = @_;
4440 my $ret = mon_cmd
($vmid, "device_del", id
=> $deviceid);
4443 sub qemu_iothread_add
{
4444 my ($vmid, $deviceid, $device) = @_;
4446 if ($device->{iothread
}) {
4447 my $iothreads = vm_iothreads_list
($vmid);
4448 qemu_objectadd
($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4452 sub qemu_iothread_del
{
4453 my ($vmid, $deviceid, $device) = @_;
4455 if ($device->{iothread
}) {
4456 my $iothreads = vm_iothreads_list
($vmid);
4457 qemu_objectdel
($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4461 sub qemu_objectadd
{
4462 my ($vmid, $objectid, $qomtype) = @_;
4464 mon_cmd
($vmid, "object-add", id
=> $objectid, "qom-type" => $qomtype);
4469 sub qemu_objectdel
{
4470 my ($vmid, $objectid) = @_;
4472 mon_cmd
($vmid, "object-del", id
=> $objectid);
4478 my ($storecfg, $vmid, $device) = @_;
4480 my $kvmver = get_running_qemu_version
($vmid);
4481 my $io_uring = min_version
($kvmver, 6, 0);
4482 my $drive = print_drive_commandline_full
($storecfg, $vmid, $device, undef, $io_uring);
4483 $drive =~ s/\\/\\\\/g;
4484 my $ret = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "drive_add auto \"$drive\"");
4486 # If the command succeeds qemu prints: "OK
"
4487 return 1 if $ret =~ m/OK/s;
4489 die "adding drive failed
: $ret\n";
4493 my ($vmid, $deviceid) = @_;
4495 my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-
$deviceid");
4498 return 1 if $ret eq "";
4500 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4501 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4503 die "deleting drive
$deviceid failed
: $ret\n";
4506 sub qemu_deviceaddverify {
4507 my ($vmid, $deviceid) = @_;
4509 for (my $i = 0; $i <= 5; $i++) {
4510 my $devices_list = vm_devices_list($vmid);
4511 return 1 if defined($devices_list->{$deviceid});
4515 die "error on hotplug device
'$deviceid'\n";
4519 sub qemu_devicedelverify {
4520 my ($vmid, $deviceid) = @_;
4522 # need to verify that the device is correctly removed as device_del
4523 # is async and empty return is not reliable
4525 for (my $i = 0; $i <= 5; $i++) {
4526 my $devices_list = vm_devices_list($vmid);
4527 return 1 if !defined($devices_list->{$deviceid});
4531 die "error on hot-unplugging device
'$deviceid'\n";
4534 sub qemu_findorcreatescsihw {
4535 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4537 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4539 my $scsihwid="$controller_prefix$controller";
4540 my $devices_list = vm_devices_list($vmid);
4542 if (!defined($devices_list->{$scsihwid})) {
4543 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4549 sub qemu_deletescsihw {
4550 my ($conf, $vmid, $opt) = @_;
4552 my $device = parse_drive($opt, $conf->{$opt});
4554 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4555 vm_deviceunplug($vmid, $conf, "virtioscsi
$device->{index}");
4559 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4561 my $devices_list = vm_devices_list($vmid);
4562 foreach my $opt (keys %{$devices_list}) {
4563 if (is_valid_drivename($opt)) {
4564 my $drive = parse_drive($opt, $conf->{$opt});
4565 if ($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4571 my $scsihwid="scsihw
$controller";
4573 vm_deviceunplug($vmid, $conf, $scsihwid);
4578 sub qemu_add_pci_bridge {
4579 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4585 print_pci_addr($device, $bridges, $arch, $machine_type);
4587 while (my ($k, $v) = each %$bridges) {
4590 return 1 if !defined($bridgeid) || $bridgeid < 1;
4592 my $bridge = "pci
.$bridgeid";
4593 my $devices_list = vm_devices_list($vmid);
4595 if (!defined($devices_list->{$bridge})) {
4596 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4602 sub qemu_set_link_status {
4603 my ($vmid, $device, $up) = @_;
4605 mon_cmd($vmid, "set_link
", name => $device,
4606 up => $up ? JSON::true : JSON::false);
4609 sub qemu_netdevadd {
4610 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4612 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4613 my %options = split(/[=,]/, $netdev);
4615 if (defined(my $vhost = $options{vhost})) {
4616 $options{vhost} = JSON::boolean(PVE::JSONSchema::parse_boolean($vhost));
4619 if (defined(my $queues = $options{queues})) {
4620 $options{queues} = $queues + 0;
4623 mon_cmd($vmid, "netdev_add
", %options);
4627 sub qemu_netdevdel {
4628 my ($vmid, $deviceid) = @_;
4630 mon_cmd($vmid, "netdev_del
", id => $deviceid);
4633 sub qemu_usb_hotplug {
4634 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4638 # remove the old one first
4639 vm_deviceunplug($vmid, $conf, $deviceid);
4641 # check if xhci controller is necessary and available
4642 my $devicelist = vm_devices_list($vmid);
4644 if (!$devicelist->{xhci}) {
4645 my $pciaddr = print_pci_addr("xhci
", undef, $arch, $machine_type);
4646 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_qemu_xhci_controller($pciaddr));
4649 # print_usbdevice_full expects the parsed device
4650 my $d = parse_usb_device($device->{host});
4651 $d->{usb3} = $device->{usb3};
4654 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4657 sub qemu_cpu_hotplug {
4658 my ($vmid, $conf, $vcpus) = @_;
4660 my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4663 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4664 $sockets = $conf->{sockets} if $conf->{sockets};
4665 my $cores = $conf->{cores} || 1;
4666 my $maxcpus = $sockets * $cores;
4668 $vcpus = $maxcpus if !$vcpus;
4670 die "you can
't add more vcpus than maxcpus\n"
4671 if $vcpus > $maxcpus;
4673 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4675 if ($vcpus < $currentvcpus) {
4677 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4679 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4680 qemu_devicedel($vmid, "cpu$i");
4682 my $currentrunningvcpus = undef;
4684 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4685 last if scalar(@{$currentrunningvcpus}) == $i-1;
4686 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4690 #update conf after each succesfull cpu unplug
4691 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4692 PVE::QemuConfig->write_config($vmid, $conf);
4695 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4701 my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4702 die "vcpus in running vm does not match its configuration\n"
4703 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4705 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4707 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4708 my $cpustr = print_cpu_device($conf, $i);
4709 qemu_deviceadd($vmid, $cpustr);
4712 my $currentrunningvcpus = undef;
4714 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4715 last if scalar(@{$currentrunningvcpus}) == $i;
4716 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4720 #update conf after each succesfull cpu hotplug
4721 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4722 PVE::QemuConfig->write_config($vmid, $conf);
4726 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4727 mon_cmd($vmid, "cpu-add", id => int($i));
4732 sub qemu_block_set_io_throttle {
4733 my ($vmid, $deviceid,
4734 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4735 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4736 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4737 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4739 return if !check_running($vmid) ;
4741 mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4743 bps_rd => int($bps_rd),
4744 bps_wr => int($bps_wr),
4746 iops_rd => int($iops_rd),
4747 iops_wr => int($iops_wr),
4748 bps_max => int($bps_max),
4749 bps_rd_max => int($bps_rd_max),
4750 bps_wr_max => int($bps_wr_max),
4751 iops_max => int($iops_max),
4752 iops_rd_max => int($iops_rd_max),
4753 iops_wr_max => int($iops_wr_max),
4754 bps_max_length => int($bps_max_length),
4755 bps_rd_max_length => int($bps_rd_max_length),
4756 bps_wr_max_length => int($bps_wr_max_length),
4757 iops_max_length => int($iops_max_length),
4758 iops_rd_max_length => int($iops_rd_max_length),
4759 iops_wr_max_length => int($iops_wr_max_length),
4764 sub qemu_block_resize {
4765 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4767 my $running = check_running($vmid);
4769 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4771 return if !$running;
4773 my $padding = (1024 - $size % 1024) % 1024;
4774 $size = $size + $padding;
4779 device => $deviceid,
4785 sub qemu_volume_snapshot {
4786 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4788 my $running = check_running($vmid);
4790 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4791 mon_cmd($vmid, 'blockdev-snapshot-internal-sync
', device => $deviceid, name => $snap);
4793 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4797 sub qemu_volume_snapshot_delete {
4798 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4800 my $running = check_running($vmid);
4805 my $conf = PVE::QemuConfig->load_config($vmid);
4806 PVE::QemuConfig->foreach_volume($conf, sub {
4807 my ($ds, $drive) = @_;
4808 $running = 1 if $drive->{file} eq $volid;
4812 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4813 mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync
', device => $deviceid, name => $snap);
4815 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4819 sub set_migration_caps {
4820 my ($vmid, $savevm) = @_;
4822 my $qemu_support = eval { mon_cmd($vmid, "query-proxmox-support") };
4824 my $bitmap_prop = $savevm ? 'pbs-dirty-bitmap-savevm
' : 'pbs-dirty-bitmap-migration
';
4825 my $dirty_bitmaps = $qemu_support->{$bitmap_prop} ? 1 : 0;
4830 "auto-converge" => 1,
4832 "x-rdma-pin-all" => 0,
4835 "dirty-bitmaps" => $dirty_bitmaps,
4838 my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4840 for my $supported_capability (@$supported_capabilities) {
4842 capability => $supported_capability->{capability},
4843 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4847 mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4851 my ($conf, $func, @param) = @_;
4855 my $test_volid = sub {
4856 my ($key, $drive, $snapname) = @_;
4858 my $volid = $drive->{file};
4861 $volhash->{$volid}->{cdrom} //= 1;
4862 $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4864 my $replicate = $drive->{replicate} // 1;
4865 $volhash->{$volid}->{replicate} //= 0;
4866 $volhash->{$volid}->{replicate} = 1 if $replicate;
4868 $volhash->{$volid}->{shared} //= 0;
4869 $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4871 $volhash->{$volid}->{referenced_in_config} //= 0;
4872 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
4874 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4875 if defined($snapname);
4877 my $size = $drive->{size};
4878 $volhash->{$volid}->{size} //= $size if $size;
4880 $volhash->{$volid}->{is_vmstate} //= 0;
4881 $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate
';
4883 $volhash->{$volid}->{is_tpmstate} //= 0;
4884 $volhash->{$volid}->{is_tpmstate} = 1 if $key eq 'tpmstate0
';
4886 $volhash->{$volid}->{is_unused} //= 0;
4887 $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4889 $volhash->{$volid}->{drivename} = $key if is_valid_drivename($key);
4892 my $include_opts = {
4893 extra_keys => ['vmstate
'],
4894 include_unused => 1,
4897 PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4898 foreach my $snapname (keys %{$conf->{snapshots}}) {
4899 my $snap = $conf->{snapshots}->{$snapname};
4900 PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4903 foreach my $volid (keys %$volhash) {
4904 &$func($volid, $volhash->{$volid}, @param);
4908 my $fast_plug_option = {
4916 'vmstatestorage
' => 1,
4921 for my $opt (keys %$confdesc_cloudinit) {
4922 $fast_plug_option->{$opt} = 1;
4925 # hotplug changes in [PENDING]
4926 # $selection hash can be used to only apply specified options, for
4927 # example: { cores => 1 } (only apply changed 'cores
')
4928 # $errors ref is used to return error messages
4929 sub vmconfig_hotplug_pending {
4930 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4932 my $defaults = load_defaults();
4933 my $arch = get_vm_arch($conf);
4934 my $machine_type = get_vm_machine($conf, undef, $arch);
4936 # commit values which do not have any impact on running VM first
4937 # Note: those option cannot raise errors, we we do not care about
4938 # $selection and always apply them.
4940 my $add_error = sub {
4941 my ($opt, $msg) = @_;
4942 $errors->{$opt} = "hotplug problem - $msg";
4945 my $cloudinit_pending_properties = PVE::QemuServer::cloudinit_pending_properties();
4947 my $cloudinit_record_changed = sub {
4948 my ($conf, $opt, $old, $new) = @_;
4949 return if !$cloudinit_pending_properties->{$opt};
4951 my $ci = ($conf->{cloudinit} //= {});
4953 my $recorded = $ci->{$opt};
4954 my %added = map { $_ => 1 } PVE::Tools::split_list(delete($ci->{added}) // '');
4956 if (defined($new)) {
4957 if (defined($old)) {
4958 # an existing value is being modified
4959 if (defined($recorded)) {
4960 # the value was already not in sync
4961 if ($new eq $recorded) {
4962 # a value is being reverted to the cloud-init state:
4964 delete $added{$opt};
4966 # the value was changed multiple times, do nothing
4968 } elsif ($added{$opt}) {
4969 # the value had been marked as added and is being changed, do nothing
4971 # the value is new, record it:
4975 # a new value is being added
4976 if (defined($recorded)) {
4977 # it was already not in sync
4978 if ($new eq $recorded) {
4979 # a value is being reverted to the cloud-init state:
4981 delete $added{$opt};
4983 # the value had temporarily been removed, do nothing
4985 } elsif ($added{$opt}) {
4986 # the value had been marked as added already, do nothing
4988 # the value is new, add it
4992 } elsif (!defined($old)) {
4993 # a non-existent value is being removed? ignore...
4995 # a value is being deleted
4996 if (defined($recorded)) {
4997 # a value was already recorded, just keep it
4998 } elsif ($added{$opt}) {
4999 # the value was marked as added, remove it
5000 delete $added{$opt};
5002 # a previously unrecorded value is being removed, record the old value:
5007 my $added = join(',', sort keys %added);
5008 $ci->{added} = $added if length($added);
5012 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5013 if ($fast_plug_option->{$opt}) {
5014 my $new = delete $conf->{pending}->{$opt};
5015 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $new);
5016 $conf->{$opt} = $new;
5022 PVE::QemuConfig->write_config($vmid, $conf);
5025 my $ostype = $conf->{ostype};
5026 my $version = extract_version($machine_type, get_running_qemu_version($vmid));
5027 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
5028 my $usb_hotplug = $hotplug_features->{usb}
5029 && min_version($version, 7, 1)
5030 && defined($ostype) && ($ostype eq 'l26
' || windows_version($ostype) > 7);
5032 my $cgroup = PVE::QemuServer::CGroup->new($vmid);
5033 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
5035 foreach my $opt (sort keys %$pending_delete_hash) {
5036 next if $selection && !$selection->{$opt};
5037 my $force = $pending_delete_hash->{$opt}->{force};
5039 if ($opt eq 'hotplug
') {
5040 die "skip\n" if ($conf->{hotplug} =~ /memory/);
5041 } elsif ($opt eq 'tablet
') {
5042 die "skip\n" if !$hotplug_features->{usb};
5043 if ($defaults->{tablet}) {
5044 vm_deviceplug($storecfg, $conf, $vmid, 'tablet
', $arch, $machine_type);
5045 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard
', $arch, $machine_type)
5046 if $arch eq 'aarch64
';
5048 vm_deviceunplug($vmid, $conf, 'tablet
');
5049 vm_deviceunplug($vmid, $conf, 'keyboard
') if $arch eq 'aarch64
';
5051 } elsif ($opt =~ m/^usb(\d+)$/) {
5053 die "skip\n" if !$usb_hotplug;
5054 vm_deviceunplug($vmid, $conf, "usbredirdev$index"); # if it's a spice port
5055 vm_deviceunplug
($vmid, $conf, $opt);
5056 } elsif ($opt eq 'vcpus') {
5057 die "skip\n" if !$hotplug_features->{cpu
};
5058 qemu_cpu_hotplug
($vmid, $conf, undef);
5059 } elsif ($opt eq 'balloon') {
5060 # enable balloon device is not hotpluggable
5061 die "skip\n" if defined($conf->{balloon
}) && $conf->{balloon
} == 0;
5062 # here we reset the ballooning value to memory
5063 my $balloon = $conf->{memory
} || $defaults->{memory
};
5064 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5065 } elsif ($fast_plug_option->{$opt}) {
5067 } elsif ($opt =~ m/^net(\d+)$/) {
5068 die "skip\n" if !$hotplug_features->{network
};
5069 vm_deviceunplug
($vmid, $conf, $opt);
5070 } elsif (is_valid_drivename
($opt)) {
5071 die "skip\n" if !$hotplug_features->{disk
} || $opt =~ m/(ide|sata)(\d+)/;
5072 vm_deviceunplug
($vmid, $conf, $opt);
5073 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5074 } elsif ($opt =~ m/^memory$/) {
5075 die "skip\n" if !$hotplug_features->{memory
};
5076 PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults);
5077 } elsif ($opt eq 'cpuunits') {
5078 $cgroup->change_cpu_shares(undef);
5079 } elsif ($opt eq 'cpulimit') {
5080 $cgroup->change_cpu_quota(undef, undef); # reset, cgroup module can better decide values
5086 &$add_error($opt, $err) if $err ne "skip\n";
5088 my $old = delete $conf->{$opt};
5089 $cloudinit_record_changed->($conf, $opt, $old, undef);
5090 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5095 foreach my $opt (keys %{$conf->{pending
}}) {
5096 next if $selection && !$selection->{$opt};
5097 my $value = $conf->{pending
}->{$opt};
5099 if ($opt eq 'hotplug') {
5100 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug
} =~ /memory/);
5101 } elsif ($opt eq 'tablet') {
5102 die "skip\n" if !$hotplug_features->{usb
};
5104 vm_deviceplug
($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5105 vm_deviceplug
($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5106 if $arch eq 'aarch64';
5107 } elsif ($value == 0) {
5108 vm_deviceunplug
($vmid, $conf, 'tablet');
5109 vm_deviceunplug
($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5111 } elsif ($opt =~ m/^usb(\d+)$/) {
5113 die "skip\n" if !$usb_hotplug;
5114 my $d = eval { parse_property_string
($usbdesc->{format
}, $value) };
5116 if ($d->{host
} eq 'spice') {
5117 $id = "usbredirdev$index";
5119 qemu_usb_hotplug
($storecfg, $conf, $vmid, $id, $d, $arch, $machine_type);
5120 } elsif ($opt eq 'vcpus') {
5121 die "skip\n" if !$hotplug_features->{cpu
};
5122 qemu_cpu_hotplug
($vmid, $conf, $value);
5123 } elsif ($opt eq 'balloon') {
5124 # enable/disable balloning device is not hotpluggable
5125 my $old_balloon_enabled = !!(!defined($conf->{balloon
}) || $conf->{balloon
});
5126 my $new_balloon_enabled = !!(!defined($conf->{pending
}->{balloon
}) || $conf->{pending
}->{balloon
});
5127 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5129 # allow manual ballooning if shares is set to zero
5130 if ((defined($conf->{shares
}) && ($conf->{shares
} == 0))) {
5131 my $balloon = $conf->{pending
}->{balloon
} || $conf->{memory
} || $defaults->{memory
};
5132 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5134 } elsif ($opt =~ m/^net(\d+)$/) {
5135 # some changes can be done without hotplug
5136 vmconfig_update_net
($storecfg, $conf, $hotplug_features->{network
},
5137 $vmid, $opt, $value, $arch, $machine_type);
5138 } elsif (is_valid_drivename
($opt)) {
5139 die "skip\n" if $opt eq 'efidisk0' || $opt eq 'tpmstate0';
5140 # some changes can be done without hotplug
5141 my $drive = parse_drive
($opt, $value);
5142 if (drive_is_cloudinit
($drive)) {
5143 $cloudinit_opt = [$opt, $drive];
5144 # apply all the other changes first, then generate the cloudinit disk
5147 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5148 $vmid, $opt, $value, $arch, $machine_type);
5149 } elsif ($opt =~ m/^memory$/) { #dimms
5150 die "skip\n" if !$hotplug_features->{memory
};
5151 $value = PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults, $value);
5152 } elsif ($opt eq 'cpuunits') {
5153 my $new_cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{pending
}->{$opt}); #clamp
5154 $cgroup->change_cpu_shares($new_cpuunits);
5155 } elsif ($opt eq 'cpulimit') {
5156 my $cpulimit = $conf->{pending
}->{$opt} == 0 ?
-1 : int($conf->{pending
}->{$opt} * 100000);
5157 $cgroup->change_cpu_quota($cpulimit, 100000);
5158 } elsif ($opt eq 'agent') {
5159 vmconfig_update_agent
($conf, $opt, $value);
5161 die "skip\n"; # skip non-hot-pluggable options
5165 &$add_error($opt, $err) if $err ne "skip\n";
5167 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $value);
5168 $conf->{$opt} = $value;
5169 delete $conf->{pending
}->{$opt};
5173 if (defined($cloudinit_opt)) {
5174 my ($opt, $drive) = @$cloudinit_opt;
5175 my $value = $conf->{pending
}->{$opt};
5177 my $temp = {%$conf, $opt => $value};
5178 PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($temp, $vmid);
5179 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5180 $vmid, $opt, $value, $arch, $machine_type);
5183 &$add_error($opt, $err) if $err ne "skip\n";
5185 $conf->{$opt} = $value;
5186 delete $conf->{pending
}->{$opt};
5190 # unplug xhci controller if no usb device is left
5193 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
5194 next if !defined($conf->{"usb$i"});
5199 vm_deviceunplug
($vmid, $conf, 'xhci');
5203 PVE
::QemuConfig-
>write_config($vmid, $conf);
5205 if ($hotplug_features->{cloudinit
} && PVE
::QemuServer
::Cloudinit
::has_changes
($conf)) {
5206 PVE
::QemuServer
::vmconfig_update_cloudinit_drive
($storecfg, $conf, $vmid);
5210 sub try_deallocate_drive
{
5211 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5213 if (($force || $key =~ /^unused/) && !drive_is_cdrom
($drive, 1)) {
5214 my $volid = $drive->{file
};
5215 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
5216 my $sid = PVE
::Storage
::parse_volume_id
($volid);
5217 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5219 # check if the disk is really unused
5220 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5221 if PVE
::QemuServer
::Drive
::is_volume_in_use
($storecfg, $conf, $key, $volid);
5222 PVE
::Storage
::vdisk_free
($storecfg, $volid);
5225 # If vm is not owner of this disk remove from config
5233 sub vmconfig_delete_or_detach_drive
{
5234 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5236 my $drive = parse_drive
($opt, $conf->{$opt});
5238 my $rpcenv = PVE
::RPCEnvironment
::get
();
5239 my $authuser = $rpcenv->get_user();
5242 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5243 try_deallocate_drive
($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5245 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $drive);
5251 sub vmconfig_apply_pending
{
5252 my ($vmid, $conf, $storecfg, $errors, $skip_cloud_init) = @_;
5254 return if !scalar(keys %{$conf->{pending
}});
5256 my $add_apply_error = sub {
5257 my ($opt, $msg) = @_;
5258 my $err_msg = "unable to apply pending change $opt : $msg";
5259 $errors->{$opt} = $err_msg;
5265 my $pending_delete_hash = PVE
::QemuConfig-
>parse_pending_delete($conf->{pending
}->{delete});
5266 foreach my $opt (sort keys %$pending_delete_hash) {
5267 my $force = $pending_delete_hash->{$opt}->{force
};
5269 if ($opt =~ m/^unused/) {
5270 die "internal error";
5271 } elsif (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5272 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5276 $add_apply_error->($opt, $err);
5278 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5279 delete $conf->{$opt};
5283 PVE
::QemuConfig-
>cleanup_pending($conf);
5285 my $generate_cloudinit = $skip_cloud_init ?
0 : undef;
5287 foreach my $opt (keys %{$conf->{pending
}}) { # add/change
5288 next if $opt eq 'delete'; # just to be sure
5290 if (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5291 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, parse_drive
($opt, $conf->{$opt}))
5295 $add_apply_error->($opt, $err);
5298 if (is_valid_drivename
($opt)) {
5299 my $drive = parse_drive
($opt, $conf->{pending
}->{$opt});
5300 $generate_cloudinit //= 1 if drive_is_cloudinit
($drive);
5303 $conf->{$opt} = delete $conf->{pending
}->{$opt};
5307 # write all changes at once to avoid unnecessary i/o
5308 PVE
::QemuConfig-
>write_config($vmid, $conf);
5309 if ($generate_cloudinit) {
5310 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5311 # After successful generation and if there were changes to be applied, update the
5312 # config to drop the {cloudinit} entry.
5313 PVE
::QemuConfig-
>write_config($vmid, $conf);
5318 sub vmconfig_update_net
{
5319 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5321 my $newnet = parse_net
($value);
5323 if ($conf->{$opt}) {
5324 my $oldnet = parse_net
($conf->{$opt});
5326 if (safe_string_ne
($oldnet->{model
}, $newnet->{model
}) ||
5327 safe_string_ne
($oldnet->{macaddr
}, $newnet->{macaddr
}) ||
5328 safe_num_ne
($oldnet->{queues
}, $newnet->{queues
}) ||
5329 safe_num_ne
($oldnet->{mtu
}, $newnet->{mtu
}) ||
5330 !($newnet->{bridge
} && $oldnet->{bridge
})) { # bridge/nat mode change
5332 # for non online change, we try to hot-unplug
5333 die "skip\n" if !$hotplug;
5334 vm_deviceunplug
($vmid, $conf, $opt);
5337 die "internal error" if $opt !~ m/net(\d+)/;
5338 my $iface = "tap${vmid}i$1";
5340 if (safe_string_ne
($oldnet->{bridge
}, $newnet->{bridge
}) ||
5341 safe_num_ne
($oldnet->{tag
}, $newnet->{tag
}) ||
5342 safe_string_ne
($oldnet->{trunks
}, $newnet->{trunks
}) ||
5343 safe_num_ne
($oldnet->{firewall
}, $newnet->{firewall
})) {
5344 PVE
::Network
::tap_unplug
($iface);
5347 PVE
::Network
::SDN
::Zones
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5349 PVE
::Network
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5351 } elsif (safe_num_ne
($oldnet->{rate
}, $newnet->{rate
})) {
5352 # Rate can be applied on its own but any change above needs to
5353 # include the rate in tap_plug since OVS resets everything.
5354 PVE
::Network
::tap_rate_limit
($iface, $newnet->{rate
});
5357 if (safe_string_ne
($oldnet->{link_down
}, $newnet->{link_down
})) {
5358 qemu_set_link_status
($vmid, $opt, !$newnet->{link_down
});
5366 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5372 sub vmconfig_update_agent
{
5373 my ($conf, $opt, $value) = @_;
5375 die "skip\n" if !$conf->{$opt};
5377 my $hotplug_options = { fstrim_cloned_disks
=> 1 };
5379 my $old_agent = parse_guest_agent
($conf);
5380 my $agent = parse_guest_agent
({$opt => $value});
5382 for my $option (keys %$agent) { # added/changed options
5383 next if defined($hotplug_options->{$option});
5384 die "skip\n" if safe_string_ne
($agent->{$option}, $old_agent->{$option});
5387 for my $option (keys %$old_agent) { # removed options
5388 next if defined($hotplug_options->{$option});
5389 die "skip\n" if safe_string_ne
($old_agent->{$option}, $agent->{$option});
5392 return; # either no actual change (e.g., format string reordered) or just hotpluggable changes
5395 sub vmconfig_update_disk
{
5396 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5398 my $drive = parse_drive
($opt, $value);
5400 if ($conf->{$opt} && (my $old_drive = parse_drive
($opt, $conf->{$opt}))) {
5401 my $media = $drive->{media
} || 'disk';
5402 my $oldmedia = $old_drive->{media
} || 'disk';
5403 die "unable to change media type\n" if $media ne $oldmedia;
5405 if (!drive_is_cdrom
($old_drive)) {
5407 if ($drive->{file
} ne $old_drive->{file
}) {
5409 die "skip\n" if !$hotplug;
5411 # unplug and register as unused
5412 vm_deviceunplug
($vmid, $conf, $opt);
5413 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive)
5416 # update existing disk
5418 # skip non hotpluggable value
5419 if (safe_string_ne
($drive->{aio
}, $old_drive->{aio
}) ||
5420 safe_string_ne
($drive->{discard
}, $old_drive->{discard
}) ||
5421 safe_string_ne
($drive->{iothread
}, $old_drive->{iothread
}) ||
5422 safe_string_ne
($drive->{queues
}, $old_drive->{queues
}) ||
5423 safe_string_ne
($drive->{cache
}, $old_drive->{cache
}) ||
5424 safe_string_ne
($drive->{ssd
}, $old_drive->{ssd
}) ||
5425 safe_string_ne
($drive->{ro
}, $old_drive->{ro
})) {
5430 if (safe_num_ne
($drive->{mbps
}, $old_drive->{mbps
}) ||
5431 safe_num_ne
($drive->{mbps_rd
}, $old_drive->{mbps_rd
}) ||
5432 safe_num_ne
($drive->{mbps_wr
}, $old_drive->{mbps_wr
}) ||
5433 safe_num_ne
($drive->{iops
}, $old_drive->{iops
}) ||
5434 safe_num_ne
($drive->{iops_rd
}, $old_drive->{iops_rd
}) ||
5435 safe_num_ne
($drive->{iops_wr
}, $old_drive->{iops_wr
}) ||
5436 safe_num_ne
($drive->{mbps_max
}, $old_drive->{mbps_max
}) ||
5437 safe_num_ne
($drive->{mbps_rd_max
}, $old_drive->{mbps_rd_max
}) ||
5438 safe_num_ne
($drive->{mbps_wr_max
}, $old_drive->{mbps_wr_max
}) ||
5439 safe_num_ne
($drive->{iops_max
}, $old_drive->{iops_max
}) ||
5440 safe_num_ne
($drive->{iops_rd_max
}, $old_drive->{iops_rd_max
}) ||
5441 safe_num_ne
($drive->{iops_wr_max
}, $old_drive->{iops_wr_max
}) ||
5442 safe_num_ne
($drive->{bps_max_length
}, $old_drive->{bps_max_length
}) ||
5443 safe_num_ne
($drive->{bps_rd_max_length
}, $old_drive->{bps_rd_max_length
}) ||
5444 safe_num_ne
($drive->{bps_wr_max_length
}, $old_drive->{bps_wr_max_length
}) ||
5445 safe_num_ne
($drive->{iops_max_length
}, $old_drive->{iops_max_length
}) ||
5446 safe_num_ne
($drive->{iops_rd_max_length
}, $old_drive->{iops_rd_max_length
}) ||
5447 safe_num_ne
($drive->{iops_wr_max_length
}, $old_drive->{iops_wr_max_length
})) {
5449 qemu_block_set_io_throttle
(
5451 ($drive->{mbps
} || 0)*1024*1024,
5452 ($drive->{mbps_rd
} || 0)*1024*1024,
5453 ($drive->{mbps_wr
} || 0)*1024*1024,
5454 $drive->{iops
} || 0,
5455 $drive->{iops_rd
} || 0,
5456 $drive->{iops_wr
} || 0,
5457 ($drive->{mbps_max
} || 0)*1024*1024,
5458 ($drive->{mbps_rd_max
} || 0)*1024*1024,
5459 ($drive->{mbps_wr_max
} || 0)*1024*1024,
5460 $drive->{iops_max
} || 0,
5461 $drive->{iops_rd_max
} || 0,
5462 $drive->{iops_wr_max
} || 0,
5463 $drive->{bps_max_length
} || 1,
5464 $drive->{bps_rd_max_length
} || 1,
5465 $drive->{bps_wr_max_length
} || 1,
5466 $drive->{iops_max_length
} || 1,
5467 $drive->{iops_rd_max_length
} || 1,
5468 $drive->{iops_wr_max_length
} || 1,
5478 if ($drive->{file
} eq 'none') {
5479 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5480 if (drive_is_cloudinit
($old_drive)) {
5481 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive);
5484 my $path = get_iso_path
($storecfg, $vmid, $drive->{file
});
5486 # force eject if locked
5487 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5490 mon_cmd
($vmid, "blockdev-change-medium",
5491 id
=> "$opt", filename
=> "$path");
5499 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5501 PVE
::Storage
::activate_volumes
($storecfg, [$drive->{file
}]) if $drive->{file
} !~ m
|^/dev/.+|;
5502 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5505 sub vmconfig_update_cloudinit_drive
{
5506 my ($storecfg, $conf, $vmid) = @_;
5508 my $cloudinit_ds = undef;
5509 my $cloudinit_drive = undef;
5511 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5512 my ($ds, $drive) = @_;
5513 if (PVE
::QemuServer
::drive_is_cloudinit
($drive)) {
5514 $cloudinit_ds = $ds;
5515 $cloudinit_drive = $drive;
5519 return if !$cloudinit_drive;
5521 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5522 PVE
::QemuConfig-
>write_config($vmid, $conf);
5525 my $running = PVE
::QemuServer
::check_running
($vmid);
5528 my $path = PVE
::Storage
::path
($storecfg, $cloudinit_drive->{file
});
5530 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$cloudinit_ds");
5531 mon_cmd
($vmid, "blockdev-change-medium", id
=> "$cloudinit_ds", filename
=> "$path");
5536 # called in locked context by incoming migration
5537 sub vm_migrate_get_nbd_disks
{
5538 my ($storecfg, $conf, $replicated_volumes) = @_;
5540 my $local_volumes = {};
5541 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5542 my ($ds, $drive) = @_;
5544 return if drive_is_cdrom
($drive);
5545 return if $ds eq 'tpmstate0';
5547 my $volid = $drive->{file
};
5551 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
5553 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5554 return if $scfg->{shared
};
5556 # replicated disks re-use existing state via bitmap
5557 my $use_existing = $replicated_volumes->{$volid} ?
1 : 0;
5558 $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
5560 return $local_volumes;
5563 # called in locked context by incoming migration
5564 sub vm_migrate_alloc_nbd_disks
{
5565 my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
5568 foreach my $opt (sort keys %$source_volumes) {
5569 my ($volid, $storeid, $volname, $drive, $use_existing, $format) = @{$source_volumes->{$opt}};
5571 if ($use_existing) {
5572 $nbd->{$opt}->{drivestr
} = print_drive
($drive);
5573 $nbd->{$opt}->{volid
} = $volid;
5574 $nbd->{$opt}->{replicated
} = 1;
5578 # storage mapping + volname = regular migration
5579 # storage mapping + format = remote migration
5580 # order of precedence, filtered by whether storage supports it:
5581 # 1. explicit requested format
5582 # 2. format of current volume
5583 # 3. default format of storage
5584 if (!$storagemap->{identity
}) {
5585 $storeid = PVE
::JSONSchema
::map_id
($storagemap, $storeid);
5586 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
5587 if (!$format || !grep { $format eq $_ } @$validFormats) {
5589 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5590 my $fileFormat = qemu_img_format
($scfg, $volname);
5591 $format = $fileFormat
5592 if grep { $fileFormat eq $_ } @$validFormats;
5594 $format //= $defFormat;
5597 # can't happen for remote migration, so $volname is always defined
5598 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5599 $format = qemu_img_format
($scfg, $volname);
5602 my $size = $drive->{size
} / 1024;
5603 my $newvolid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $format, undef, $size);
5604 my $newdrive = $drive;
5605 $newdrive->{format
} = $format;
5606 $newdrive->{file
} = $newvolid;
5607 my $drivestr = print_drive
($newdrive);
5608 $nbd->{$opt}->{drivestr
} = $drivestr;
5609 $nbd->{$opt}->{volid
} = $newvolid;
5615 # see vm_start_nolock for parameters, additionally:
5617 # storagemap = parsed storage map for allocating NBD disks
5619 my ($storecfg, $vmid, $params, $migrate_opts) = @_;
5621 return PVE
::QemuConfig-
>lock_config($vmid, sub {
5622 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migrate_opts->{migratedfrom
});
5624 die "you can't start a vm if it's a template\n"
5625 if !$params->{skiptemplate
} && PVE
::QemuConfig-
>is_template($conf);
5627 my $has_suspended_lock = PVE
::QemuConfig-
>has_lock($conf, 'suspended');
5628 my $has_backup_lock = PVE
::QemuConfig-
>has_lock($conf, 'backup');
5630 my $running = check_running
($vmid, undef, $migrate_opts->{migratedfrom
});
5632 if ($has_backup_lock && $running) {
5633 # a backup is currently running, attempt to start the guest in the
5634 # existing QEMU instance
5635 return vm_resume
($vmid);
5638 PVE
::QemuConfig-
>check_lock($conf)
5639 if !($params->{skiplock
} || $has_suspended_lock);
5641 $params->{resume
} = $has_suspended_lock || defined($conf->{vmstate
});
5643 die "VM $vmid already running\n" if $running;
5645 if (my $storagemap = $migrate_opts->{storagemap
}) {
5646 my $replicated = $migrate_opts->{replicated_volumes
};
5647 my $disks = vm_migrate_get_nbd_disks
($storecfg, $conf, $replicated);
5648 $migrate_opts->{nbd
} = vm_migrate_alloc_nbd_disks
($storecfg, $vmid, $disks, $storagemap);
5650 foreach my $opt (keys %{$migrate_opts->{nbd
}}) {
5651 $conf->{$opt} = $migrate_opts->{nbd
}->{$opt}->{drivestr
};
5655 return vm_start_nolock
($storecfg, $vmid, $conf, $params, $migrate_opts);
5661 # statefile => 'tcp', 'unix' for migration or path/volid for RAM state
5662 # skiplock => 0/1, skip checking for config lock
5663 # skiptemplate => 0/1, skip checking whether VM is template
5664 # forcemachine => to force QEMU machine (rollback/migration)
5665 # forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
5666 # timeout => in seconds
5667 # paused => start VM in paused state (backup)
5668 # resume => resume from hibernation
5679 # nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
5680 # migratedfrom => source node
5681 # spice_ticket => used for spice migration, passed via tunnel/stdin
5682 # network => CIDR of migration network
5683 # type => secure/insecure - tunnel over encrypted connection or plain-text
5684 # nbd_proto_version => int, 0 for TCP, 1 for UNIX
5685 # replicated_volumes => which volids should be re-used with bitmaps for nbd migration
5686 # offline_volumes => new volids of offline migrated disks like tpmstate and cloudinit, not yet
5687 # contained in config
5688 sub vm_start_nolock
{
5689 my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
5691 my $statefile = $params->{statefile
};
5692 my $resume = $params->{resume
};
5694 my $migratedfrom = $migrate_opts->{migratedfrom
};
5695 my $migration_type = $migrate_opts->{type
};
5699 # clean up leftover reboot request files
5700 eval { clear_reboot_request
($vmid); };
5703 if (!$statefile && scalar(keys %{$conf->{pending
}})) {
5704 vmconfig_apply_pending
($vmid, $conf, $storecfg);
5705 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5708 # don't regenerate the ISO if the VM is started as part of a live migration
5709 # this way we can reuse the old ISO with the correct config
5710 if (!$migratedfrom) {
5711 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5712 # FIXME: apply_cloudinit_config updates $conf in this case, and it would only drop
5713 # $conf->{cloudinit}, so we could just not do this?
5714 # But we do it above, so for now let's be consistent.
5715 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5719 # override offline migrated volumes, conf is out of date still
5720 if (my $offline_volumes = $migrate_opts->{offline_volumes
}) {
5721 for my $key (sort keys $offline_volumes->%*) {
5722 my $parsed = parse_drive
($key, $conf->{$key});
5723 $parsed->{file
} = $offline_volumes->{$key};
5724 $conf->{$key} = print_drive
($parsed);
5728 my $defaults = load_defaults
();
5730 # set environment variable useful inside network script
5731 # for remote migration the config is available on the target node!
5732 if (!$migrate_opts->{remote_node
}) {
5733 $ENV{PVE_MIGRATED_FROM
} = $migratedfrom;
5736 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-start', 1);
5738 my $forcemachine = $params->{forcemachine
};
5739 my $forcecpu = $params->{forcecpu
};
5741 # enforce machine and CPU type on suspended vm to ensure HW compatibility
5742 $forcemachine = $conf->{runningmachine
};
5743 $forcecpu = $conf->{runningcpu
};
5744 print "Resuming suspended VM\n";
5747 my ($cmd, $vollist, $spice_port) = config_to_command
($storecfg, $vmid,
5748 $conf, $defaults, $forcemachine, $forcecpu, $params->{'pbs-backing'});
5751 my $get_migration_ip = sub {
5752 my ($nodename) = @_;
5754 return $migration_ip if defined($migration_ip);
5756 my $cidr = $migrate_opts->{network
};
5758 if (!defined($cidr)) {
5759 my $dc_conf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5760 $cidr = $dc_conf->{migration
}->{network
};
5763 if (defined($cidr)) {
5764 my $ips = PVE
::Network
::get_local_ip_from_cidr
($cidr);
5766 die "could not get IP: no address configured on local " .
5767 "node for network '$cidr'\n" if scalar(@$ips) == 0;
5769 die "could not get IP: multiple addresses configured on local " .
5770 "node for network '$cidr'\n" if scalar(@$ips) > 1;
5772 $migration_ip = @$ips[0];
5775 $migration_ip = PVE
::Cluster
::remote_node_ip
($nodename, 1)
5776 if !defined($migration_ip);
5778 return $migration_ip;
5782 if ($statefile eq 'tcp') {
5783 my $migrate = $res->{migrate
} = { proto
=> 'tcp' };
5784 $migrate->{addr
} = "localhost";
5785 my $datacenterconf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5786 my $nodename = nodename
();
5788 if (!defined($migration_type)) {
5789 if (defined($datacenterconf->{migration
}->{type
})) {
5790 $migration_type = $datacenterconf->{migration
}->{type
};
5792 $migration_type = 'secure';
5796 if ($migration_type eq 'insecure') {
5797 $migrate->{addr
} = $get_migration_ip->($nodename);
5798 $migrate->{addr
} = "[$migrate->{addr}]" if Net
::IP
::ip_is_ipv6
($migrate->{addr
});
5801 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5802 $migrate->{port
} = PVE
::Tools
::next_migrate_port
($pfamily);
5803 $migrate->{uri
} = "tcp:$migrate->{addr}:$migrate->{port}";
5804 push @$cmd, '-incoming', $migrate->{uri
};
5807 } elsif ($statefile eq 'unix') {
5808 # should be default for secure migrations as a ssh TCP forward
5809 # tunnel is not deterministic reliable ready and fails regurarly
5810 # to set up in time, so use UNIX socket forwards
5811 my $migrate = $res->{migrate
} = { proto
=> 'unix' };
5812 $migrate->{addr
} = "/run/qemu-server/$vmid.migrate";
5813 unlink $migrate->{addr
};
5815 $migrate->{uri
} = "unix:$migrate->{addr}";
5816 push @$cmd, '-incoming', $migrate->{uri
};
5819 } elsif (-e
$statefile) {
5820 push @$cmd, '-loadstate', $statefile;
5822 my $statepath = PVE
::Storage
::path
($storecfg, $statefile);
5823 push @$vollist, $statefile;
5824 push @$cmd, '-loadstate', $statepath;
5826 } elsif ($params->{paused
}) {
5830 my $start_timeout = $params->{timeout
} // config_aware_timeout
($conf, $resume);
5832 my $pci_devices = {}; # host pci devices
5833 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
5834 my $dev = $conf->{"hostpci$i"} or next;
5835 $pci_devices->{$i} = parse_hostpci
($dev);
5838 # do not reserve pciid for mediated devices, sysfs will error out for duplicate assignment
5839 my $real_pci_devices = [ grep { !(defined($_->{mdev
}) && scalar($_->{pciid
}->@*) == 1) } values $pci_devices->%* ];
5841 # map to a flat list of pci ids
5842 my $pci_id_list = [ map { $_->{id
} } map { $_->{pciid
}->@* } $real_pci_devices->@* ];
5844 # reserve all PCI IDs before actually doing anything with them
5845 PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, $start_timeout);
5849 for my $id (sort keys %$pci_devices) {
5850 my $d = $pci_devices->{$id};
5851 for my $dev ($d->{pciid
}->@*) {
5852 my $info = PVE
::QemuServer
::PCI
::prepare_pci_device
($vmid, $dev->{id
}, $id, $d->{mdev
});
5854 # nvidia grid needs the qemu parameter '-uuid' set
5855 # use smbios uuid or mdev uuid as fallback for that
5856 if ($d->{mdev
} && !defined($uuid) && $info->{vendor
} eq '10de') {
5857 if (defined($conf->{smbios1
})) {
5858 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
5859 $uuid = $smbios_conf->{uuid
} if defined($smbios_conf->{uuid
});
5861 $uuid = PVE
::QemuServer
::PCI
::generate_mdev_uuid
($vmid, $id) if !defined($uuid);
5865 push @$cmd, '-uuid', $uuid if defined($uuid);
5868 eval { cleanup_pci_devices
($vmid, $conf) };
5873 PVE
::Storage
::activate_volumes
($storecfg, $vollist);
5876 run_command
(['/bin/systemctl', 'stop', "$vmid.scope"], outfunc
=> sub{}, errfunc
=> sub{});
5878 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5879 # timeout should be more than enough here...
5880 PVE
::Systemd
::wait_for_unit_removed
("$vmid.scope", 20);
5882 my $cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{cpuunits
});
5885 timeout
=> $statefile ?
undef : $start_timeout,
5890 # when migrating, prefix QEMU output so other side can pick up any
5891 # errors that might occur and show the user
5892 if ($migratedfrom) {
5893 $run_params{quiet
} = 1;
5894 $run_params{logfunc
} = sub { print "QEMU: $_[0]\n" };
5897 my %systemd_properties = (
5898 Slice
=> 'qemu.slice',
5899 KillMode
=> 'process',
5901 TimeoutStopUSec
=> ULONG_MAX
, # infinity
5904 if (PVE
::CGroup
::cgroup_mode
() == 2) {
5905 $systemd_properties{CPUWeight
} = $cpuunits;
5907 $systemd_properties{CPUShares
} = $cpuunits;
5910 if (my $cpulimit = $conf->{cpulimit
}) {
5911 $systemd_properties{CPUQuota
} = int($cpulimit * 100);
5913 $systemd_properties{timeout
} = 10 if $statefile; # setting up the scope shoul be quick
5915 my $run_qemu = sub {
5916 PVE
::Tools
::run_fork
sub {
5917 PVE
::Systemd
::enter_systemd_scope
($vmid, "Proxmox VE VM $vmid", %systemd_properties);
5920 if (my $tpm = $conf->{tpmstate0
}) {
5921 # start the TPM emulator so QEMU can connect on start
5922 $tpmpid = start_swtpm
($storecfg, $vmid, $tpm, $migratedfrom);
5925 my $exitcode = run_command
($cmd, %run_params);
5928 warn "stopping swtpm instance (pid $tpmpid) due to QEMU startup error\n";
5929 kill 'TERM', $tpmpid;
5931 die "QEMU exited with code $exitcode\n";
5936 if ($conf->{hugepages
}) {
5939 my $hotplug_features =
5940 parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
5941 my $hugepages_topology =
5942 PVE
::QemuServer
::Memory
::hugepages_topology
($conf, $hotplug_features->{memory
});
5944 my $hugepages_host_topology = PVE
::QemuServer
::Memory
::hugepages_host_topology
();
5946 PVE
::QemuServer
::Memory
::hugepages_mount
();
5947 PVE
::QemuServer
::Memory
::hugepages_allocate
($hugepages_topology, $hugepages_host_topology);
5949 eval { $run_qemu->() };
5951 PVE
::QemuServer
::Memory
::hugepages_reset
($hugepages_host_topology)
5952 if !$conf->{keephugepages
};
5956 PVE
::QemuServer
::Memory
::hugepages_pre_deallocate
($hugepages_topology)
5957 if !$conf->{keephugepages
};
5959 eval { PVE
::QemuServer
::Memory
::hugepages_update_locked
($code); };
5962 eval { $run_qemu->() };
5966 # deactivate volumes if start fails
5967 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
5969 eval { cleanup_pci_devices
($vmid, $conf) };
5972 die "start failed: $err";
5975 # re-reserve all PCI IDs now that we can know the actual VM PID
5976 my $pid = PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
5977 eval { PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, undef, $pid) };
5980 if (defined($res->{migrate
})) {
5981 print "migration listens on $res->{migrate}->{uri}\n";
5982 } elsif ($statefile) {
5983 eval { mon_cmd
($vmid, "cont"); };
5987 #start nbd server for storage migration
5988 if (my $nbd = $migrate_opts->{nbd
}) {
5989 my $nbd_protocol_version = $migrate_opts->{nbd_proto_version
} // 0;
5991 my $migrate_storage_uri;
5992 # nbd_protocol_version > 0 for unix socket support
5993 if ($nbd_protocol_version > 0 && ($migration_type eq 'secure' || $migration_type eq 'websocket')) {
5994 my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
5995 mon_cmd
($vmid, "nbd-server-start", addr
=> { type
=> 'unix', data
=> { path
=> $socket_path } } );
5996 $migrate_storage_uri = "nbd:unix:$socket_path";
5997 $res->{migrate
}->{unix_sockets
} = [$socket_path];
5999 my $nodename = nodename
();
6000 my $localip = $get_migration_ip->($nodename);
6001 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
6002 my $storage_migrate_port = PVE
::Tools
::next_migrate_port
($pfamily);
6004 mon_cmd
($vmid, "nbd-server-start", addr
=> {
6007 host
=> "${localip}",
6008 port
=> "${storage_migrate_port}",
6011 $localip = "[$localip]" if Net
::IP
::ip_is_ipv6
($localip);
6012 $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
6015 my $block_info = mon_cmd
($vmid, "query-block");
6016 $block_info = { map { $_->{device
} => $_ } $block_info->@* };
6018 foreach my $opt (sort keys %$nbd) {
6019 my $drivestr = $nbd->{$opt}->{drivestr
};
6020 my $volid = $nbd->{$opt}->{volid
};
6022 my $block_node = $block_info->{"drive-$opt"}->{inserted
}->{'node-name'};
6028 'node-name' => $block_node,
6029 writable
=> JSON
::true
,
6031 name
=> "drive-$opt", # NBD export name
6034 my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
6035 print "storage migration listens on $nbd_uri volume:$drivestr\n";
6036 print "re-using replicated volume: $opt - $volid\n"
6037 if $nbd->{$opt}->{replicated
};
6039 $res->{drives
}->{$opt} = $nbd->{$opt};
6040 $res->{drives
}->{$opt}->{nbd_uri
} = $nbd_uri;
6044 if ($migratedfrom) {
6046 set_migration_caps
($vmid);
6051 print "spice listens on port $spice_port\n";
6052 $res->{spice_port
} = $spice_port;
6053 if ($migrate_opts->{spice_ticket
}) {
6054 mon_cmd
($vmid, "set_password", protocol
=> 'spice', password
=>
6055 $migrate_opts->{spice_ticket
});
6056 mon_cmd
($vmid, "expire_password", protocol
=> 'spice', time => "+30");
6061 mon_cmd
($vmid, "balloon", value
=> $conf->{balloon
}*1024*1024)
6062 if !$statefile && $conf->{balloon
};
6064 foreach my $opt (keys %$conf) {
6065 next if $opt !~ m/^net\d+$/;
6066 my $nicconf = parse_net
($conf->{$opt});
6067 qemu_set_link_status
($vmid, $opt, 0) if $nicconf->{link_down
};
6069 add_nets_bridge_fdb
($conf, $vmid);
6072 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
6077 path
=> "machine/peripheral/balloon0",
6078 property
=> "guest-stats-polling-interval",
6082 log_warn
("could not set polling interval for ballooning - $@") if $@;
6086 print "Resumed VM, removing state\n";
6087 if (my $vmstate = $conf->{vmstate
}) {
6088 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6089 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6091 delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
6092 PVE
::QemuConfig-
>write_config($vmid, $conf);
6095 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'post-start');
6100 sub vm_commandline
{
6101 my ($storecfg, $vmid, $snapname) = @_;
6103 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6105 my ($forcemachine, $forcecpu);
6107 my $snapshot = $conf->{snapshots
}->{$snapname};
6108 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
6110 # check for machine or CPU overrides in snapshot
6111 $forcemachine = $snapshot->{runningmachine
};
6112 $forcecpu = $snapshot->{runningcpu
};
6114 $snapshot->{digest
} = $conf->{digest
}; # keep file digest for API
6119 my $defaults = load_defaults
();
6121 my $cmd = config_to_command
($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
6123 return PVE
::Tools
::cmd2string
($cmd);
6127 my ($vmid, $skiplock) = @_;
6129 PVE
::QemuConfig-
>lock_config($vmid, sub {
6131 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6133 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6135 mon_cmd
($vmid, "system_reset");
6139 sub get_vm_volumes
{
6143 foreach_volid
($conf, sub {
6144 my ($volid, $attr) = @_;
6146 return if $volid =~ m
|^/|;
6148 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
6151 push @$vollist, $volid;
6157 sub cleanup_pci_devices
{
6158 my ($vmid, $conf) = @_;
6160 foreach my $key (keys %$conf) {
6161 next if $key !~ m/^hostpci(\d+)$/;
6162 my $hostpciindex = $1;
6163 my $uuid = PVE
::SysFSTools
::generate_mdev_uuid
($vmid, $hostpciindex);
6164 my $d = parse_hostpci
($conf->{$key});
6166 # NOTE: avoid PVE::SysFSTools::pci_cleanup_mdev_device as it requires PCI ID and we
6167 # don't want to break ABI just for this two liner
6168 my $dev_sysfs_dir = "/sys/bus/mdev/devices/$uuid";
6170 # some nvidia vgpu driver versions want to clean the mdevs up themselves, and error
6171 # out when we do it first. so wait for 10 seconds and then try it
6172 my $pciid = $d->{pciid
}->[0]->{id
};
6173 my $info = PVE
::SysFSTools
::pci_device_info
("$pciid");
6174 if ($info->{vendor
} eq '10de') {
6178 PVE
::SysFSTools
::file_write
("$dev_sysfs_dir/remove", "1") if -e
$dev_sysfs_dir;
6181 PVE
::QemuServer
::PCI
::remove_pci_reservation
($vmid);
6184 sub vm_stop_cleanup
{
6185 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
6190 my $vollist = get_vm_volumes
($conf);
6191 PVE
::Storage
::deactivate_volumes
($storecfg, $vollist);
6193 if (my $tpmdrive = $conf->{tpmstate0
}) {
6194 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
6195 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
6197 PVE
::Storage
::unmap_volume
($storecfg, $tpm->{file
});
6202 foreach my $ext (qw(mon qmp pid vnc qga)) {
6203 unlink "/var/run/qemu-server/${vmid}.$ext";
6206 if ($conf->{ivshmem
}) {
6207 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
6208 # just delete it for now, VMs which have this already open do not
6209 # are affected, but new VMs will get a separated one. If this
6210 # becomes an issue we either add some sort of ref-counting or just
6211 # add a "don't delete on stop" flag to the ivshmem format.
6212 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name
} // $vmid);
6215 cleanup_pci_devices
($vmid, $conf);
6217 vmconfig_apply_pending
($vmid, $conf, $storecfg) if $apply_pending_changes;
6219 warn $@ if $@; # avoid errors - just warn
6222 # call only in locked context
6224 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
6226 my $pid = check_running
($vmid, $nocheck);
6231 $conf = PVE
::QemuConfig-
>load_config($vmid);
6232 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6233 if (!defined($timeout) && $shutdown && $conf->{startup
}) {
6234 my $opts = PVE
::JSONSchema
::pve_parse_startup_order
($conf->{startup
});
6235 $timeout = $opts->{down
} if $opts->{down
};
6237 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-stop');
6242 if (defined($conf) && get_qga_key
($conf, 'enabled')) {
6243 mon_cmd
($vmid, "guest-shutdown", timeout
=> $timeout);
6245 mon_cmd
($vmid, "system_powerdown");
6248 mon_cmd
($vmid, "quit");
6254 $timeout = 60 if !defined($timeout);
6257 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6262 if ($count >= $timeout) {
6264 warn "VM still running - terminating now with SIGTERM\n";
6267 die "VM quit/powerdown failed - got timeout\n";
6270 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6274 if (!check_running
($vmid, $nocheck)) {
6275 warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
6279 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
6282 die "VM quit/powerdown failed\n";
6290 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6295 if ($count >= $timeout) {
6296 warn "VM still running - terminating now with SIGKILL\n";
6301 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6304 # Note: use $nocheck to skip tests if VM configuration file exists.
6305 # We need that when migration VMs to other nodes (files already moved)
6306 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
6308 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
6310 $force = 1 if !defined($force) && !$shutdown;
6313 my $pid = check_running
($vmid, $nocheck, $migratedfrom);
6314 kill 15, $pid if $pid;
6315 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migratedfrom);
6316 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 0);
6320 PVE
::QemuConfig-
>lock_config($vmid, sub {
6321 _do_vm_stop
($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
6326 my ($vmid, $timeout) = @_;
6328 PVE
::QemuConfig-
>lock_config($vmid, sub {
6331 # only reboot if running, as qmeventd starts it again on a stop event
6332 return if !check_running
($vmid);
6334 create_reboot_request
($vmid);
6336 my $storecfg = PVE
::Storage
::config
();
6337 _do_vm_stop
($storecfg, $vmid, undef, undef, $timeout, 1);
6341 # avoid that the next normal shutdown will be confused for a reboot
6342 clear_reboot_request
($vmid);
6348 # note: if using the statestorage parameter, the caller has to check privileges
6350 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
6357 PVE
::QemuConfig-
>lock_config($vmid, sub {
6359 $conf = PVE
::QemuConfig-
>load_config($vmid);
6361 my $is_backing_up = PVE
::QemuConfig-
>has_lock($conf, 'backup');
6362 PVE
::QemuConfig-
>check_lock($conf)
6363 if !($skiplock || $is_backing_up);
6365 die "cannot suspend to disk during backup\n"
6366 if $is_backing_up && $includestate;
6368 if ($includestate) {
6369 $conf->{lock} = 'suspending';
6370 my $date = strftime
("%Y-%m-%d", localtime(time()));
6371 $storecfg = PVE
::Storage
::config
();
6372 if (!$statestorage) {
6373 $statestorage = find_vmstate_storage
($conf, $storecfg);
6374 # check permissions for the storage
6375 my $rpcenv = PVE
::RPCEnvironment
::get
();
6376 if ($rpcenv->{type
} ne 'cli') {
6377 my $authuser = $rpcenv->get_user();
6378 $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
6383 $vmstate = PVE
::QemuConfig-
>__snapshot_save_vmstate(
6384 $vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
6385 $path = PVE
::Storage
::path
($storecfg, $vmstate);
6386 PVE
::QemuConfig-
>write_config($vmid, $conf);
6388 mon_cmd
($vmid, "stop");
6392 if ($includestate) {
6394 PVE
::Storage
::activate_volumes
($storecfg, [$vmstate]);
6397 set_migration_caps
($vmid, 1);
6398 mon_cmd
($vmid, "savevm-start", statefile
=> $path);
6400 my $state = mon_cmd
($vmid, "query-savevm");
6401 if (!$state->{status
}) {
6402 die "savevm not active\n";
6403 } elsif ($state->{status
} eq 'active') {
6406 } elsif ($state->{status
} eq 'completed') {
6407 print "State saved, quitting\n";
6409 } elsif ($state->{status
} eq 'failed' && $state->{error
}) {
6410 die "query-savevm failed with error '$state->{error}'\n"
6412 die "query-savevm returned status '$state->{status}'\n";
6418 PVE
::QemuConfig-
>lock_config($vmid, sub {
6419 $conf = PVE
::QemuConfig-
>load_config($vmid);
6421 # cleanup, but leave suspending lock, to indicate something went wrong
6423 mon_cmd
($vmid, "savevm-end");
6424 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6425 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6426 delete $conf->@{qw(vmstate runningmachine runningcpu)};
6427 PVE
::QemuConfig-
>write_config($vmid, $conf);
6433 die "lock changed unexpectedly\n"
6434 if !PVE
::QemuConfig-
>has_lock($conf, 'suspending');
6436 mon_cmd
($vmid, "quit");
6437 $conf->{lock} = 'suspended';
6438 PVE
::QemuConfig-
>write_config($vmid, $conf);
6443 # $nocheck is set when called as part of a migration - in this context the
6444 # location of the config file (source or target node) is not deterministic,
6445 # since migration cannot wait for pmxcfs to process the rename
6447 my ($vmid, $skiplock, $nocheck) = @_;
6449 PVE
::QemuConfig-
>lock_config($vmid, sub {
6450 my $res = mon_cmd
($vmid, 'query-status');
6451 my $resume_cmd = 'cont';
6455 $conf = eval { PVE
::QemuConfig-
>load_config($vmid) }; # try on target node
6457 my $vmlist = PVE
::Cluster
::get_vmlist
();
6458 if (exists($vmlist->{ids
}->{$vmid})) {
6459 my $node = $vmlist->{ids
}->{$vmid}->{node
};
6460 $conf = eval { PVE
::QemuConfig-
>load_config($vmid, $node) }; # try on source node
6463 PVE
::Cluster
::cfs_update
(); # vmlist was wrong, invalidate cache
6464 $conf = PVE
::QemuConfig-
>load_config($vmid); # last try on target node again
6468 $conf = PVE
::QemuConfig-
>load_config($vmid);
6471 if ($res->{status
}) {
6472 return if $res->{status
} eq 'running'; # job done, go home
6473 $resume_cmd = 'system_wakeup' if $res->{status
} eq 'suspended';
6474 $reset = 1 if $res->{status
} eq 'shutdown';
6478 PVE
::QemuConfig-
>check_lock($conf)
6479 if !($skiplock || PVE
::QemuConfig-
>has_lock($conf, 'backup'));
6483 # required if a VM shuts down during a backup and we get a resume
6484 # request before the backup finishes for example
6485 mon_cmd
($vmid, "system_reset");
6488 add_nets_bridge_fdb
($conf, $vmid) if $resume_cmd eq 'cont';
6490 mon_cmd
($vmid, $resume_cmd);
6495 my ($vmid, $skiplock, $key) = @_;
6497 PVE
::QemuConfig-
>lock_config($vmid, sub {
6499 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6501 # there is no qmp command, so we use the human monitor command
6502 my $res = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "sendkey $key");
6503 die $res if $res ne '';
6507 # vzdump restore implementaion
6509 sub tar_archive_read_firstfile
{
6510 my $archive = shift;
6512 die "ERROR: file '$archive' does not exist\n" if ! -f
$archive;
6514 # try to detect archive type first
6515 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6516 die "unable to open file '$archive'\n";
6517 my $firstfile = <$fh>;
6521 die "ERROR: archive contaions no data\n" if !$firstfile;
6527 sub tar_restore_cleanup
{
6528 my ($storecfg, $statfile) = @_;
6530 print STDERR
"starting cleanup\n";
6532 if (my $fd = IO
::File-
>new($statfile, "r")) {
6533 while (defined(my $line = <$fd>)) {
6534 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6537 if ($volid =~ m
|^/|) {
6538 unlink $volid || die 'unlink failed\n';
6540 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6542 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6544 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6546 print STDERR
"unable to parse line in statfile - $line";
6553 sub restore_file_archive
{
6554 my ($archive, $vmid, $user, $opts) = @_;
6556 return restore_vma_archive
($archive, $vmid, $user, $opts)
6559 my $info = PVE
::Storage
::archive_info
($archive);
6560 my $format = $opts->{format
} // $info->{format
};
6561 my $comp = $info->{compression
};
6563 # try to detect archive format
6564 if ($format eq 'tar') {
6565 return restore_tar_archive
($archive, $vmid, $user, $opts);
6567 return restore_vma_archive
($archive, $vmid, $user, $opts, $comp);
6571 # hepler to remove disks that will not be used after restore
6572 my $restore_cleanup_oldconf = sub {
6573 my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
6575 my $kept_disks = {};
6577 PVE
::QemuConfig-
>foreach_volume($oldconf, sub {
6578 my ($ds, $drive) = @_;
6580 return if drive_is_cdrom
($drive, 1);
6582 my $volid = $drive->{file
};
6583 return if !$volid || $volid =~ m
|^/|;
6585 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
6586 return if !$path || !$owner || ($owner != $vmid);
6588 # Note: only delete disk we want to restore
6589 # other volumes will become unused
6590 if ($virtdev_hash->{$ds}) {
6591 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
6596 $kept_disks->{$volid} = 1;
6600 # after the restore we have no snapshots anymore
6601 for my $snapname (keys $oldconf->{snapshots
}->%*) {
6602 my $snap = $oldconf->{snapshots
}->{$snapname};
6603 if ($snap->{vmstate
}) {
6604 eval { PVE
::Storage
::vdisk_free
($storecfg, $snap->{vmstate
}); };
6610 for my $volid (keys $kept_disks->%*) {
6611 eval { PVE
::Storage
::volume_snapshot_delete
($storecfg, $volid, $snapname); };
6617 # Helper to parse vzdump backup device hints
6619 # $rpcenv: Environment, used to ckeck storage permissions
6620 # $user: User ID, to check storage permissions
6621 # $storecfg: Storage configuration
6622 # $fh: the file handle for reading the configuration
6623 # $devinfo: should contain device sizes for all backu-up'ed devices
6624 # $options: backup options (pool, default storage)
6626 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
6627 my $parse_backup_hints = sub {
6628 my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
6630 my $check_storage = sub { # assert if an image can be allocate
6631 my ($storeid, $scfg) = @_;
6632 die "Content type 'images' is not available on storage '$storeid'\n"
6633 if !$scfg->{content
}->{images
};
6634 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace'])
6635 if $user ne 'root@pam';
6638 my $virtdev_hash = {};
6639 while (defined(my $line = <$fh>)) {
6640 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6641 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6642 die "archive does not contain data for drive '$virtdev'\n"
6643 if !$devinfo->{$devname};
6645 if (defined($options->{storage
})) {
6646 $storeid = $options->{storage
} || 'local';
6647 } elsif (!$storeid) {
6650 $format = 'raw' if !$format;
6651 $devinfo->{$devname}->{devname
} = $devname;
6652 $devinfo->{$devname}->{virtdev
} = $virtdev;
6653 $devinfo->{$devname}->{format
} = $format;
6654 $devinfo->{$devname}->{storeid
} = $storeid;
6656 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6657 $check_storage->($storeid, $scfg); # permission and content type check
6659 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6660 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6662 my $drive = parse_drive
($virtdev, $2);
6664 if (drive_is_cloudinit
($drive)) {
6665 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
6666 $storeid = $options->{storage
} if defined ($options->{storage
});
6667 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6668 my $format = qemu_img_format
($scfg, $volname); # has 'raw' fallback
6670 $check_storage->($storeid, $scfg); # permission and content type check
6672 $virtdev_hash->{$virtdev} = {
6674 storeid
=> $storeid,
6675 size
=> PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
,
6682 return $virtdev_hash;
6685 # Helper to allocate and activate all volumes required for a restore
6687 # $storecfg: Storage configuration
6688 # $virtdev_hash: as returned by parse_backup_hints()
6690 # Returns: { $virtdev => $volid }
6691 my $restore_allocate_devices = sub {
6692 my ($storecfg, $virtdev_hash, $vmid) = @_;
6695 foreach my $virtdev (sort keys %$virtdev_hash) {
6696 my $d = $virtdev_hash->{$virtdev};
6697 my $alloc_size = int(($d->{size
} + 1024 - 1)/1024);
6698 my $storeid = $d->{storeid
};
6699 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6701 # test if requested format is supported
6702 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
6703 my $supported = grep { $_ eq $d->{format
} } @$validFormats;
6704 $d->{format
} = $defFormat if !$supported;
6707 if ($d->{is_cloudinit
}) {
6708 $name = "vm-$vmid-cloudinit";
6709 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6710 if ($scfg->{path
}) {
6711 $name .= ".$d->{format}";
6715 my $volid = PVE
::Storage
::vdisk_alloc
(
6716 $storecfg, $storeid, $vmid, $d->{format
}, $name, $alloc_size);
6718 print STDERR
"new volume ID is '$volid'\n";
6719 $d->{volid
} = $volid;
6721 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
6723 $map->{$virtdev} = $volid;
6729 sub restore_update_config_line
{
6730 my ($cookie, $map, $line, $unique) = @_;
6732 return '' if $line =~ m/^\#qmdump\#/;
6733 return '' if $line =~ m/^\#vzdump\#/;
6734 return '' if $line =~ m/^lock:/;
6735 return '' if $line =~ m/^unused\d+:/;
6736 return '' if $line =~ m/^parent:/;
6740 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
6741 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6742 # try to convert old 1.X settings
6743 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6744 foreach my $devconfig (PVE
::Tools
::split_list
($ethcfg)) {
6745 my ($model, $macaddr) = split(/\=/, $devconfig);
6746 $macaddr = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if !$macaddr || $unique;
6749 bridge
=> "vmbr$ind",
6750 macaddr
=> $macaddr,
6752 my $netstr = print_net
($net);
6754 $res .= "net$cookie->{netcount}: $netstr\n";
6755 $cookie->{netcount
}++;
6757 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6758 my ($id, $netstr) = ($1, $2);
6759 my $net = parse_net
($netstr);
6760 $net->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if $net->{macaddr
};
6761 $netstr = print_net
($net);
6762 $res .= "$id: $netstr\n";
6763 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk|tpmstate)\d+):\s*(\S+)\s*$/) {
6766 my $di = parse_drive
($virtdev, $value);
6767 if (defined($di->{backup
}) && !$di->{backup
}) {
6769 } elsif ($map->{$virtdev}) {
6770 delete $di->{format
}; # format can change on restore
6771 $di->{file
} = $map->{$virtdev};
6772 $value = print_drive
($di);
6773 $res .= "$virtdev: $value\n";
6777 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6779 if ($vmgenid ne '0') {
6780 # always generate a new vmgenid if there was a valid one setup
6781 $vmgenid = generate_uuid
();
6783 $res .= "vmgenid: $vmgenid\n";
6784 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6785 my ($uuid, $uuid_str);
6786 UUID
::generate
($uuid);
6787 UUID
::unparse
($uuid, $uuid_str);
6788 my $smbios1 = parse_smbios1
($2);
6789 $smbios1->{uuid
} = $uuid_str;
6790 $res .= $1.print_smbios1
($smbios1)."\n";
6798 my $restore_deactivate_volumes = sub {
6799 my ($storecfg, $virtdev_hash) = @_;
6802 for my $dev (values $virtdev_hash->%*) {
6803 push $vollist->@*, $dev->{volid
} if $dev->{volid
};
6806 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
6807 print STDERR
$@ if $@;
6810 my $restore_destroy_volumes = sub {
6811 my ($storecfg, $virtdev_hash) = @_;
6813 for my $dev (values $virtdev_hash->%*) {
6814 my $volid = $dev->{volid
} or next;
6816 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6817 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6819 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6823 my $restore_merge_config = sub {
6824 my ($filename, $backup_conf_raw, $override_conf) = @_;
6826 my $backup_conf = parse_vm_config
($filename, $backup_conf_raw);
6827 for my $key (keys $override_conf->%*) {
6828 $backup_conf->{$key} = $override_conf->{$key};
6831 return $backup_conf;
6835 my ($cfg, $vmid) = @_;
6837 my $info = PVE
::Storage
::vdisk_list
($cfg, undef, $vmid, undef, 'images');
6839 my $volid_hash = {};
6840 foreach my $storeid (keys %$info) {
6841 foreach my $item (@{$info->{$storeid}}) {
6842 next if !($item->{volid
} && $item->{size
});
6843 $item->{path
} = PVE
::Storage
::path
($cfg, $item->{volid
});
6844 $volid_hash->{$item->{volid
}} = $item;
6851 sub update_disk_config
{
6852 my ($vmid, $conf, $volid_hash) = @_;
6855 my $prefix = "VM $vmid";
6857 # used and unused disks
6858 my $referenced = {};
6860 # Note: it is allowed to define multiple storages with same path (alias), so
6861 # we need to check both 'volid' and real 'path' (two different volid can point
6862 # to the same path).
6864 my $referencedpath = {};
6867 PVE
::QemuConfig-
>foreach_volume($conf, sub {
6868 my ($opt, $drive) = @_;
6870 my $volid = $drive->{file
};
6872 my $volume = $volid_hash->{$volid};
6874 # mark volid as "in-use" for next step
6875 $referenced->{$volid} = 1;
6876 if ($volume && (my $path = $volume->{path
})) {
6877 $referencedpath->{$path} = 1;
6880 return if drive_is_cdrom
($drive);
6883 my ($updated, $msg) = PVE
::QemuServer
::Drive
::update_disksize
($drive, $volume->{size
});
6884 if (defined($updated)) {
6886 $conf->{$opt} = print_drive
($updated);
6887 print "$prefix ($opt): $msg\n";
6891 # remove 'unusedX' entry if volume is used
6892 PVE
::QemuConfig-
>foreach_unused_volume($conf, sub {
6893 my ($opt, $drive) = @_;
6895 my $volid = $drive->{file
};
6899 $path = $volid_hash->{$volid}->{path
} if $volid_hash->{$volid};
6900 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6901 print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
6903 delete $conf->{$opt};
6906 $referenced->{$volid} = 1;
6907 $referencedpath->{$path} = 1 if $path;
6910 foreach my $volid (sort keys %$volid_hash) {
6911 next if $volid =~ m/vm-$vmid-state-/;
6912 next if $referenced->{$volid};
6913 my $path = $volid_hash->{$volid}->{path
};
6914 next if !$path; # just to be sure
6915 next if $referencedpath->{$path};
6917 my $key = PVE
::QemuConfig-
>add_unused_volume($conf, $volid);
6918 print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
6919 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6926 my ($vmid, $nolock, $dryrun) = @_;
6928 my $cfg = PVE
::Storage
::config
();
6930 print "rescan volumes...\n";
6931 my $volid_hash = scan_volids
($cfg, $vmid);
6933 my $updatefn = sub {
6936 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6938 PVE
::QemuConfig-
>check_lock($conf);
6941 foreach my $volid (keys %$volid_hash) {
6942 my $info = $volid_hash->{$volid};
6943 $vm_volids->{$volid} = $info if $info->{vmid
} && $info->{vmid
} == $vmid;
6946 my $changes = update_disk_config
($vmid, $conf, $vm_volids);
6948 PVE
::QemuConfig-
>write_config($vmid, $conf) if $changes && !$dryrun;
6951 if (defined($vmid)) {
6955 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6958 my $vmlist = config_list
();
6959 foreach my $vmid (keys %$vmlist) {
6963 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6969 sub restore_proxmox_backup_archive
{
6970 my ($archive, $vmid, $user, $options) = @_;
6972 my $storecfg = PVE
::Storage
::config
();
6974 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($archive);
6975 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6977 my $fingerprint = $scfg->{fingerprint
};
6978 my $keyfile = PVE
::Storage
::PBSPlugin
::pbs_encryption_key_file_name
($storecfg, $storeid);
6980 my $repo = PVE
::PBSClient
::get_repository
($scfg);
6981 my $namespace = $scfg->{namespace
};
6983 # This is only used for `pbs-restore` and the QEMU PBS driver (live-restore)
6984 my $password = PVE
::Storage
::PBSPlugin
::pbs_get_password
($scfg, $storeid);
6985 local $ENV{PBS_PASSWORD
} = $password;
6986 local $ENV{PBS_FINGERPRINT
} = $fingerprint if defined($fingerprint);
6988 my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6989 PVE
::Storage
::parse_volname
($storecfg, $archive);
6991 die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
6993 die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
6995 my $tmpdir = "/var/tmp/vzdumptmp$$";
6999 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7000 # disable interrupts (always do cleanups)
7004 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7006 # Note: $oldconf is undef if VM does not exists
7007 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7008 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7009 my $new_conf_raw = '';
7011 my $rpcenv = PVE
::RPCEnvironment
::get
();
7012 my $devinfo = {}; # info about drives included in backup
7013 my $virtdev_hash = {}; # info about allocated drives
7021 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7023 my $cfgfn = "$tmpdir/qemu-server.conf";
7024 my $firewall_config_fn = "$tmpdir/fw.conf";
7025 my $index_fn = "$tmpdir/index.json";
7027 my $cmd = "restore";
7029 my $param = [$pbs_backup_name, "index.json", $index_fn];
7030 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7031 my $index = PVE
::Tools
::file_get_contents
($index_fn);
7032 $index = decode_json
($index);
7034 foreach my $info (@{$index->{files
}}) {
7035 if ($info->{filename
} =~ m/^(drive-\S+).img.fidx$/) {
7037 if ($info->{size
} =~ m/^(\d+)$/) { # untaint size
7038 $devinfo->{$devname}->{size
} = $1;
7040 die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
7045 my $is_qemu_server_backup = scalar(
7046 grep { $_->{filename
} eq 'qemu-server.conf.blob' } @{$index->{files
}}
7048 if (!$is_qemu_server_backup) {
7049 die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
7051 my $has_firewall_config = scalar(grep { $_->{filename
} eq 'fw.conf.blob' } @{$index->{files
}});
7053 $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
7054 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7056 if ($has_firewall_config) {
7057 $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
7058 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7060 my $pve_firewall_dir = '/etc/pve/firewall';
7061 mkdir $pve_firewall_dir; # make sure the dir exists
7062 PVE
::Tools
::file_copy
($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
7065 my $fh = IO
::File-
>new($cfgfn, "r") ||
7066 die "unable to read qemu-server.conf - $!\n";
7068 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
7070 # fixme: rate limit?
7072 # create empty/temp config
7073 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\nlock: create");
7075 $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
7078 my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
7080 foreach my $virtdev (sort keys %$virtdev_hash) {
7081 my $d = $virtdev_hash->{$virtdev};
7082 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7084 # this fails if storage is unavailable
7085 my $volid = $d->{volid
};
7086 my $path = PVE
::Storage
::path
($storecfg, $volid);
7088 # for live-restore we only want to preload the efidisk and TPM state
7089 next if $options->{live
} && $virtdev ne 'efidisk0' && $virtdev ne 'tpmstate0';
7092 if (defined(my $ns = $scfg->{namespace
})) {
7093 @ns_arg = ('--ns', $ns);
7096 my $pbs_restore_cmd = [
7097 '/usr/bin/pbs-restore',
7098 '--repository', $repo,
7101 "$d->{devname}.img.fidx",
7106 push @$pbs_restore_cmd, '--format', $d->{format
} if $d->{format
};
7107 push @$pbs_restore_cmd, '--keyfile', $keyfile if -e
$keyfile;
7109 if (PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $volid)) {
7110 push @$pbs_restore_cmd, '--skip-zero';
7113 my $dbg_cmdstring = PVE
::Tools
::cmd2string
($pbs_restore_cmd);
7114 print "restore proxmox backup image: $dbg_cmdstring\n";
7115 run_command
($pbs_restore_cmd);
7118 $fh->seek(0, 0) || die "seek failed - $!\n";
7120 my $cookie = { netcount
=> 0 };
7121 while (defined(my $line = <$fh>)) {
7122 $new_conf_raw .= restore_update_config_line
(
7134 if ($err || !$options->{live
}) {
7135 $restore_deactivate_volumes->($storecfg, $virtdev_hash);
7141 $restore_destroy_volumes->($storecfg, $virtdev_hash);
7145 if ($options->{live
}) {
7146 # keep lock during live-restore
7147 $new_conf_raw .= "\nlock: create";
7150 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $options->{override_conf
});
7151 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7153 eval { rescan
($vmid, 1); };
7156 PVE
::AccessControl
::add_vm_to_pool
($vmid, $options->{pool
}) if $options->{pool
};
7158 if ($options->{live
}) {
7164 local $SIG{PIPE
} = sub { die "got signal ($!) - abort\n"; };
7166 my $conf = PVE
::QemuConfig-
>load_config($vmid);
7167 die "cannot do live-restore for template\n" if PVE
::QemuConfig-
>is_template($conf);
7169 # these special drives are already restored before start
7170 delete $devinfo->{'drive-efidisk0'};
7171 delete $devinfo->{'drive-tpmstate0-backup'};
7175 keyfile
=> $keyfile,
7176 snapshot
=> $pbs_backup_name,
7177 namespace
=> $namespace,
7179 pbs_live_restore
($vmid, $conf, $storecfg, $devinfo, $pbs_opts);
7181 PVE
::QemuConfig-
>remove_lock($vmid, "create");
7185 sub pbs_live_restore
{
7186 my ($vmid, $conf, $storecfg, $restored_disks, $opts) = @_;
7188 print "starting VM for live-restore\n";
7189 print "repository: '$opts->{repo}', snapshot: '$opts->{snapshot}'\n";
7191 my $pbs_backing = {};
7192 for my $ds (keys %$restored_disks) {
7193 $ds =~ m/^drive-(.*)$/;
7195 $pbs_backing->{$confname} = {
7196 repository
=> $opts->{repo
},
7197 snapshot
=> $opts->{snapshot
},
7198 archive
=> "$ds.img.fidx",
7200 $pbs_backing->{$confname}->{keyfile
} = $opts->{keyfile
} if -e
$opts->{keyfile
};
7201 $pbs_backing->{$confname}->{namespace
} = $opts->{namespace
} if defined($opts->{namespace
});
7203 my $drive = parse_drive
($confname, $conf->{$confname});
7204 print "restoring '$ds' to '$drive->{file}'\n";
7207 my $drives_streamed = 0;
7209 # make sure HA doesn't interrupt our restore by stopping the VM
7210 if (PVE
::HA
::Config
::vm_is_ha_managed
($vmid)) {
7211 run_command
(['ha-manager', 'set', "vm:$vmid", '--state', 'started']);
7214 # start VM with backing chain pointing to PBS backup, environment vars for PBS driver
7215 # in QEMU (PBS_PASSWORD and PBS_FINGERPRINT) are already set by our caller
7216 vm_start_nolock
($storecfg, $vmid, $conf, {paused
=> 1, 'pbs-backing' => $pbs_backing}, {});
7218 my $qmeventd_fd = register_qmeventd_handle
($vmid);
7220 # begin streaming, i.e. data copy from PBS to target disk for every vol,
7221 # this will effectively collapse the backing image chain consisting of
7222 # [target <- alloc-track -> PBS snapshot] to just [target] (alloc-track
7223 # removes itself once all backing images vanish with 'auto-remove=on')
7225 for my $ds (sort keys %$restored_disks) {
7226 my $job_id = "restore-$ds";
7227 mon_cmd
($vmid, 'block-stream',
7228 'job-id' => $job_id,
7231 $jobs->{$job_id} = {};
7234 mon_cmd
($vmid, 'cont');
7235 qemu_drive_mirror_monitor
($vmid, undef, $jobs, 'auto', 0, 'stream');
7237 print "restore-drive jobs finished successfully, removing all tracking block devices"
7238 ." to disconnect from Proxmox Backup Server\n";
7240 for my $ds (sort keys %$restored_disks) {
7241 mon_cmd
($vmid, 'blockdev-del', 'node-name' => "$ds-pbs");
7244 close($qmeventd_fd);
7250 warn "An error occurred during live-restore: $err\n";
7251 _do_vm_stop
($storecfg, $vmid, 1, 1, 10, 0, 1);
7252 die "live-restore failed\n";
7256 sub restore_vma_archive
{
7257 my ($archive, $vmid, $user, $opts, $comp) = @_;
7259 my $readfrom = $archive;
7261 my $cfg = PVE
::Storage
::config
();
7263 my $bwlimit = $opts->{bwlimit
};
7265 my $dbg_cmdstring = '';
7266 my $add_pipe = sub {
7268 push @$commands, $cmd;
7269 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
7270 $dbg_cmdstring .= PVE
::Tools
::cmd2string
($cmd);
7275 if ($archive eq '-') {
7278 # If we use a backup from a PVE defined storage we also consider that
7279 # storage's rate limit:
7280 my (undef, $volid) = PVE
::Storage
::path_to_volume_id
($cfg, $archive);
7281 if (defined($volid)) {
7282 my ($sid, undef) = PVE
::Storage
::parse_volume_id
($volid);
7283 my $readlimit = PVE
::Storage
::get_bandwidth_limit
('restore', [$sid], $bwlimit);
7285 print STDERR
"applying read rate limit: $readlimit\n";
7286 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
7287 $add_pipe->($cstream);
7293 my $info = PVE
::Storage
::decompressor_info
('vma', $comp);
7294 my $cmd = $info->{decompressor
};
7295 push @$cmd, $readfrom;
7299 my $tmpdir = "/var/tmp/vzdumptmp$$";
7302 # disable interrupts (always do cleanups)
7306 local $SIG{HUP
} = sub { warn "got interrupt - ignored\n"; };
7308 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
7309 POSIX
::mkfifo
($mapfifo, 0600);
7311 my $openfifo = sub { open($fifofh, '>', $mapfifo) or die $! };
7313 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
7318 my $devinfo = {}; # info about drives included in backup
7319 my $virtdev_hash = {}; # info about allocated drives
7321 my $rpcenv = PVE
::RPCEnvironment
::get
();
7323 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7325 # Note: $oldconf is undef if VM does not exist
7326 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7327 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7328 my $new_conf_raw = '';
7332 my $print_devmap = sub {
7333 my $cfgfn = "$tmpdir/qemu-server.conf";
7335 # we can read the config - that is already extracted
7336 my $fh = IO
::File-
>new($cfgfn, "r") ||
7337 die "unable to read qemu-server.conf - $!\n";
7339 my $fwcfgfn = "$tmpdir/qemu-server.fw";
7341 my $pve_firewall_dir = '/etc/pve/firewall';
7342 mkdir $pve_firewall_dir; # make sure the dir exists
7343 PVE
::Tools
::file_copy
($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
7346 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
7348 foreach my $info (values %{$virtdev_hash}) {
7349 my $storeid = $info->{storeid
};
7350 next if defined($storage_limits{$storeid});
7352 my $limit = PVE
::Storage
::get_bandwidth_limit
('restore', [$storeid], $bwlimit) // 0;
7353 print STDERR
"rate limit for storage $storeid: $limit KiB/s\n" if $limit;
7354 $storage_limits{$storeid} = $limit * 1024;
7357 foreach my $devname (keys %$devinfo) {
7358 die "found no device mapping information for device '$devname'\n"
7359 if !$devinfo->{$devname}->{virtdev
};
7362 # create empty/temp config
7364 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\n");
7365 $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
7369 my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
7371 # print restore information to $fifofh
7372 foreach my $virtdev (sort keys %$virtdev_hash) {
7373 my $d = $virtdev_hash->{$virtdev};
7374 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7376 my $storeid = $d->{storeid
};
7377 my $volid = $d->{volid
};
7380 if (my $limit = $storage_limits{$storeid}) {
7381 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
7384 my $write_zeros = 1;
7385 if (PVE
::Storage
::volume_has_feature
($cfg, 'sparseinit', $volid)) {
7389 my $path = PVE
::Storage
::path
($cfg, $volid);
7391 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
7393 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
7396 $fh->seek(0, 0) || die "seek failed - $!\n";
7398 my $cookie = { netcount
=> 0 };
7399 while (defined(my $line = <$fh>)) {
7400 $new_conf_raw .= restore_update_config_line
(
7417 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7418 local $SIG{ALRM
} = sub { die "got timeout\n"; };
7420 $oldtimeout = alarm($timeout);
7427 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
7428 my ($dev_id, $size, $devname) = ($1, $2, $3);
7429 $devinfo->{$devname} = { size
=> $size, dev_id
=> $dev_id };
7430 } elsif ($line =~ m/^CTIME: /) {
7431 # we correctly received the vma config, so we can disable
7432 # the timeout now for disk allocation (set to 10 minutes, so
7433 # that we always timeout if something goes wrong)
7436 print $fifofh "done\n";
7437 my $tmp = $oldtimeout || 0;
7438 $oldtimeout = undef;
7445 print "restore vma archive: $dbg_cmdstring\n";
7446 run_command
($commands, input
=> $input, outfunc
=> $parser, afterfork
=> $openfifo);
7450 alarm($oldtimeout) if $oldtimeout;
7452 $restore_deactivate_volumes->($cfg, $virtdev_hash);
7454 close($fifofh) if $fifofh;
7459 $restore_destroy_volumes->($cfg, $virtdev_hash);
7463 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $opts->{override_conf
});
7464 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7466 eval { rescan
($vmid, 1); };
7469 PVE
::AccessControl
::add_vm_to_pool
($vmid, $opts->{pool
}) if $opts->{pool
};
7472 sub restore_tar_archive
{
7473 my ($archive, $vmid, $user, $opts) = @_;
7475 if (scalar(keys $opts->{override_conf
}->%*) > 0) {
7476 my $keystring = join(' ', keys $opts->{override_conf
}->%*);
7477 die "cannot pass along options ($keystring) when restoring from tar archive\n";
7480 if ($archive ne '-') {
7481 my $firstfile = tar_archive_read_firstfile
($archive);
7482 die "ERROR: file '$archive' does not look like a QemuServer vzdump backup\n"
7483 if $firstfile ne 'qemu-server.conf';
7486 my $storecfg = PVE
::Storage
::config
();
7488 # avoid zombie disks when restoring over an existing VM -> cleanup first
7489 # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
7490 # skiplock=1 because qmrestore has set the 'create' lock itself already
7491 my $vmcfgfn = PVE
::QemuConfig-
>config_file($vmid);
7492 destroy_vm
($storecfg, $vmid, 1, { lock => 'restore' }) if -f
$vmcfgfn;
7494 my $tocmd = "/usr/lib/qemu-server/qmextract";
7496 $tocmd .= " --storage " . PVE
::Tools
::shellquote
($opts->{storage
}) if $opts->{storage
};
7497 $tocmd .= " --pool " . PVE
::Tools
::shellquote
($opts->{pool
}) if $opts->{pool
};
7498 $tocmd .= ' --prealloc' if $opts->{prealloc
};
7499 $tocmd .= ' --info' if $opts->{info
};
7501 # tar option "xf" does not autodetect compression when read from STDIN,
7502 # so we pipe to zcat
7503 my $cmd = "zcat -f|tar xf " . PVE
::Tools
::shellquote
($archive) . " " .
7504 PVE
::Tools
::shellquote
("--to-command=$tocmd");
7506 my $tmpdir = "/var/tmp/vzdumptmp$$";
7509 local $ENV{VZDUMP_TMPDIR
} = $tmpdir;
7510 local $ENV{VZDUMP_VMID
} = $vmid;
7511 local $ENV{VZDUMP_USER
} = $user;
7513 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7514 my $new_conf_raw = '';
7516 # disable interrupts (always do cleanups)
7520 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7528 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7530 if ($archive eq '-') {
7531 print "extracting archive from STDIN\n";
7532 run_command
($cmd, input
=> "<&STDIN");
7534 print "extracting archive '$archive'\n";
7538 return if $opts->{info
};
7542 my $statfile = "$tmpdir/qmrestore.stat";
7543 if (my $fd = IO
::File-
>new($statfile, "r")) {
7544 while (defined (my $line = <$fd>)) {
7545 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
7546 $map->{$1} = $2 if $1;
7548 print STDERR
"unable to parse line in statfile - $line\n";
7554 my $confsrc = "$tmpdir/qemu-server.conf";
7556 my $srcfd = IO
::File-
>new($confsrc, "r") || die "unable to open file '$confsrc'\n";
7558 my $cookie = { netcount
=> 0 };
7559 while (defined (my $line = <$srcfd>)) {
7560 $new_conf_raw .= restore_update_config_line
(
7571 tar_restore_cleanup
($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info
};
7577 PVE
::Tools
::file_set_contents
($conffile, $new_conf_raw);
7579 PVE
::Cluster
::cfs_update
(); # make sure we read new file
7581 eval { rescan
($vmid, 1); };
7585 sub foreach_storage_used_by_vm
{
7586 my ($conf, $func) = @_;
7590 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7591 my ($ds, $drive) = @_;
7592 return if drive_is_cdrom
($drive);
7594 my $volid = $drive->{file
};
7596 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
7597 $sidhash->{$sid} = $sid if $sid;
7600 foreach my $sid (sort keys %$sidhash) {
7605 my $qemu_snap_storage = {
7608 sub do_snapshots_with_qemu
{
7609 my ($storecfg, $volid, $deviceid) = @_;
7611 return if $deviceid =~ m/tpmstate0/;
7613 my $storage_name = PVE
::Storage
::parse_volume_id
($volid);
7614 my $scfg = $storecfg->{ids
}->{$storage_name};
7615 die "could not find storage '$storage_name'\n" if !defined($scfg);
7617 if ($qemu_snap_storage->{$scfg->{type
}} && !$scfg->{krbd
}){
7621 if ($volid =~ m/\.(qcow2|qed)$/){
7628 sub qga_check_running
{
7629 my ($vmid, $nowarn) = @_;
7631 eval { mon_cmd
($vmid, "guest-ping", timeout
=> 3); };
7633 warn "QEMU Guest Agent is not running - $@" if !$nowarn;
7639 sub template_create
{
7640 my ($vmid, $conf, $disk) = @_;
7642 my $storecfg = PVE
::Storage
::config
();
7644 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7645 my ($ds, $drive) = @_;
7647 return if drive_is_cdrom
($drive);
7648 return if $disk && $ds ne $disk;
7650 my $volid = $drive->{file
};
7651 return if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
7653 my $voliddst = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
7654 $drive->{file
} = $voliddst;
7655 $conf->{$ds} = print_drive
($drive);
7656 PVE
::QemuConfig-
>write_config($vmid, $conf);
7660 sub convert_iscsi_path
{
7663 if ($path =~ m
|^iscsi
://([^/]+)/([^/]+)/(.+)$|) {
7668 my $initiator_name = get_initiator_name
();
7670 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
7671 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
7674 die "cannot convert iscsi path '$path', unkown format\n";
7677 sub qemu_img_convert
{
7678 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized, $bwlimit) = @_;
7680 my $storecfg = PVE
::Storage
::config
();
7681 my ($src_storeid, $src_volname) = PVE
::Storage
::parse_volume_id
($src_volid, 1);
7682 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid, 1);
7684 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
7688 my $src_is_iscsi = 0;
7692 PVE
::Storage
::activate_volumes
($storecfg, [$src_volid], $snapname);
7693 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7694 $src_format = qemu_img_format
($src_scfg, $src_volname);
7695 $src_path = PVE
::Storage
::path
($storecfg, $src_volid, $snapname);
7696 $src_is_iscsi = ($src_path =~ m
|^iscsi
://|);
7697 $cachemode = 'none' if $src_scfg->{type
} eq 'zfspool';
7698 } elsif (-f
$src_volid || -b
$src_volid) {
7699 $src_path = $src_volid;
7700 if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7705 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
7707 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7708 my $dst_format = qemu_img_format
($dst_scfg, $dst_volname);
7709 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7710 my $dst_is_iscsi = ($dst_path =~ m
|^iscsi
://|);
7713 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
7714 push @$cmd, '-l', "snapshot.name=$snapname"
7715 if $snapname && $src_format && $src_format eq "qcow2";
7716 push @$cmd, '-t', 'none' if $dst_scfg->{type
} eq 'zfspool';
7717 push @$cmd, '-T', $cachemode if defined($cachemode);
7718 push @$cmd, '-r', "${bwlimit}K" if defined($bwlimit);
7720 if ($src_is_iscsi) {
7721 push @$cmd, '--image-opts';
7722 $src_path = convert_iscsi_path
($src_path);
7723 } elsif ($src_format) {
7724 push @$cmd, '-f', $src_format;
7727 if ($dst_is_iscsi) {
7728 push @$cmd, '--target-image-opts';
7729 $dst_path = convert_iscsi_path
($dst_path);
7731 push @$cmd, '-O', $dst_format;
7734 push @$cmd, $src_path;
7736 if (!$dst_is_iscsi && $is_zero_initialized) {
7737 push @$cmd, "zeroinit:$dst_path";
7739 push @$cmd, $dst_path;
7744 if($line =~ m/\((\S+)\/100\
%\)/){
7746 my $transferred = int($size * $percent / 100);
7747 my $total_h = render_bytes
($size, 1);
7748 my $transferred_h = render_bytes
($transferred, 1);
7750 print "transferred $transferred_h of $total_h ($percent%)\n";
7755 eval { run_command
($cmd, timeout
=> undef, outfunc
=> $parser); };
7757 die "copy failed: $err" if $err;
7760 sub qemu_img_format
{
7761 my ($scfg, $volname) = @_;
7763 if ($scfg->{path
} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7770 sub qemu_drive_mirror
{
7771 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
7773 $jobs = {} if !$jobs;
7777 $jobs->{"drive-$drive"} = {};
7779 if ($dst_volid =~ /^nbd:/) {
7780 $qemu_target = $dst_volid;
7783 my $storecfg = PVE
::Storage
::config
();
7784 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid);
7786 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7788 $format = qemu_img_format
($dst_scfg, $dst_volname);
7790 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7792 $qemu_target = $is_zero_initialized ?
"zeroinit:$dst_path" : $dst_path;
7795 my $opts = { timeout
=> 10, device
=> "drive-$drive", mode
=> "existing", sync
=> "full", target
=> $qemu_target };
7796 $opts->{format
} = $format if $format;
7798 if (defined($src_bitmap)) {
7799 $opts->{sync
} = 'incremental';
7800 $opts->{bitmap
} = $src_bitmap;
7801 print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
7804 if (defined($bwlimit)) {
7805 $opts->{speed
} = $bwlimit * 1024;
7806 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7808 print "drive mirror is starting for drive-$drive\n";
7811 # if a job already runs for this device we get an error, catch it for cleanup
7812 eval { mon_cmd
($vmid, "drive-mirror", %$opts); };
7814 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7816 die "mirroring error: $err\n";
7819 qemu_drive_mirror_monitor
($vmid, $vmiddst, $jobs, $completion, $qga);
7822 # $completion can be either
7823 # 'complete': wait until all jobs are ready, block-job-complete them (default)
7824 # 'cancel': wait until all jobs are ready, block-job-cancel them
7825 # 'skip': wait until all jobs are ready, return with block jobs in ready state
7826 # 'auto': wait until all jobs disappear, only use for jobs which complete automatically
7827 sub qemu_drive_mirror_monitor
{
7828 my ($vmid, $vmiddst, $jobs, $completion, $qga, $op) = @_;
7830 $completion //= 'complete';
7834 my $err_complete = 0;
7836 my $starttime = time ();
7838 die "block job ('$op') timed out\n" if $err_complete > 300;
7840 my $stats = mon_cmd
($vmid, "query-block-jobs");
7843 my $running_jobs = {};
7844 for my $stat (@$stats) {
7845 next if $stat->{type
} ne $op;
7846 $running_jobs->{$stat->{device
}} = $stat;
7849 my $readycounter = 0;
7851 for my $job_id (sort keys %$jobs) {
7852 my $job = $running_jobs->{$job_id};
7854 my $vanished = !defined($job);
7855 my $complete = defined($jobs->{$job_id}->{complete
}) && $vanished;
7856 if($complete || ($vanished && $completion eq 'auto')) {
7857 print "$job_id: $op-job finished\n";
7858 delete $jobs->{$job_id};
7862 die "$job_id: '$op' has been cancelled\n" if !defined($job);
7864 my $busy = $job->{busy
};
7865 my $ready = $job->{ready
};
7866 if (my $total = $job->{len
}) {
7867 my $transferred = $job->{offset
} || 0;
7868 my $remaining = $total - $transferred;
7869 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7871 my $duration = $ctime - $starttime;
7872 my $total_h = render_bytes
($total, 1);
7873 my $transferred_h = render_bytes
($transferred, 1);
7875 my $status = sprintf(
7876 "transferred $transferred_h of $total_h ($percent%%) in %s",
7877 render_duration
($duration),
7882 $status .= ", still busy"; # shouldn't even happen? but mirror is weird
7884 $status .= ", ready";
7887 print "$job_id: $status\n" if !$jobs->{$job_id}->{ready
};
7888 $jobs->{$job_id}->{ready
} = $ready;
7891 $readycounter++ if $job->{ready
};
7894 last if scalar(keys %$jobs) == 0;
7896 if ($readycounter == scalar(keys %$jobs)) {
7897 print "all '$op' jobs are ready\n";
7899 # do the complete later (or has already been done)
7900 last if $completion eq 'skip' || $completion eq 'auto';
7902 if ($vmiddst && $vmiddst != $vmid) {
7903 my $agent_running = $qga && qga_check_running
($vmid);
7904 if ($agent_running) {
7905 print "freeze filesystem\n";
7906 eval { mon_cmd
($vmid, "guest-fsfreeze-freeze"); };
7909 print "suspend vm\n";
7910 eval { PVE
::QemuServer
::vm_suspend
($vmid, 1); };
7914 # if we clone a disk for a new target vm, we don't switch the disk
7915 PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs);
7917 if ($agent_running) {
7918 print "unfreeze filesystem\n";
7919 eval { mon_cmd
($vmid, "guest-fsfreeze-thaw"); };
7922 print "resume vm\n";
7923 eval { PVE
::QemuServer
::vm_resume
($vmid, 1, 1); };
7930 for my $job_id (sort keys %$jobs) {
7931 # try to switch the disk if source and destination are on the same guest
7932 print "$job_id: Completing block job_id...\n";
7935 if ($completion eq 'complete') {
7936 $op = 'block-job-complete';
7937 } elsif ($completion eq 'cancel') {
7938 $op = 'block-job-cancel';
7940 die "invalid completion value: $completion\n";
7942 eval { mon_cmd
($vmid, $op, device
=> $job_id) };
7943 if ($@ =~ m/cannot be completed/) {
7944 print "$job_id: block job cannot be completed, trying again.\n";
7947 print "$job_id: Completed successfully.\n";
7948 $jobs->{$job_id}->{complete
} = 1;
7959 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7960 die "block job ($op) error: $err";
7964 sub qemu_blockjobs_cancel
{
7965 my ($vmid, $jobs) = @_;
7967 foreach my $job (keys %$jobs) {
7968 print "$job: Cancelling block job\n";
7969 eval { mon_cmd
($vmid, "block-job-cancel", device
=> $job); };
7970 $jobs->{$job}->{cancel
} = 1;
7974 my $stats = mon_cmd
($vmid, "query-block-jobs");
7976 my $running_jobs = {};
7977 foreach my $stat (@$stats) {
7978 $running_jobs->{$stat->{device
}} = $stat;
7981 foreach my $job (keys %$jobs) {
7983 if (defined($jobs->{$job}->{cancel
}) && !defined($running_jobs->{$job})) {
7984 print "$job: Done.\n";
7985 delete $jobs->{$job};
7989 last if scalar(keys %$jobs) == 0;
7995 # Check for bug #4525: drive-mirror will open the target drive with the same aio setting as the
7996 # source, but some storages have problems with io_uring, sometimes even leading to crashes.
7997 my sub clone_disk_check_io_uring
{
7998 my ($src_drive, $storecfg, $src_storeid, $dst_storeid, $use_drive_mirror) = @_;
8000 return if !$use_drive_mirror;
8002 # Don't complain when not changing storage.
8003 # Assume if it works for the source, it'll work for the target too.
8004 return if $src_storeid eq $dst_storeid;
8006 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
8007 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
8009 my $cache_direct = drive_uses_cache_direct
($src_drive);
8011 my $src_uses_io_uring;
8012 if ($src_drive->{aio
}) {
8013 $src_uses_io_uring = $src_drive->{aio
} eq 'io_uring';
8015 $src_uses_io_uring = storage_allows_io_uring_default
($src_scfg, $cache_direct);
8018 die "target storage is known to cause issues with aio=io_uring (used by current drive)\n"
8019 if $src_uses_io_uring && !storage_allows_io_uring_default
($dst_scfg, $cache_direct);
8023 my ($storecfg, $source, $dest, $full, $newvollist, $jobs, $completion, $qga, $bwlimit) = @_;
8025 my ($vmid, $running) = $source->@{qw(vmid running)};
8026 my ($src_drivename, $drive, $snapname) = $source->@{qw(drivename drive snapname)};
8028 my ($newvmid, $dst_drivename, $efisize) = $dest->@{qw(vmid drivename efisize)};
8029 my ($storage, $format) = $dest->@{qw(storage format)};
8031 my $use_drive_mirror = $full && $running && $src_drivename && !$snapname;
8033 if ($src_drivename && $dst_drivename && $src_drivename ne $dst_drivename) {
8034 die "cloning from/to EFI disk requires EFI disk\n"
8035 if $src_drivename eq 'efidisk0' || $dst_drivename eq 'efidisk0';
8036 die "cloning from/to TPM state requires TPM state\n"
8037 if $src_drivename eq 'tpmstate0' || $dst_drivename eq 'tpmstate0';
8039 # This would lead to two device nodes in QEMU pointing to the same backing image!
8040 die "cannot change drive name when cloning disk from/to the same VM\n"
8041 if $use_drive_mirror && $vmid == $newvmid;
8044 die "cannot move TPM state while VM is running\n"
8045 if $use_drive_mirror && $src_drivename eq 'tpmstate0';
8049 print "create " . ($full ?
'full' : 'linked') . " clone of drive ";
8050 print "$src_drivename " if $src_drivename;
8051 print "($drive->{file})\n";
8054 $newvolid = PVE
::Storage
::vdisk_clone
($storecfg, $drive->{file
}, $newvmid, $snapname);
8055 push @$newvollist, $newvolid;
8057 my ($src_storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
8058 my $storeid = $storage || $src_storeid;
8060 my $dst_format = resolve_dst_disk_format
($storecfg, $storeid, $volname, $format);
8064 if (drive_is_cloudinit
($drive)) {
8065 $name = "vm-$newvmid-cloudinit";
8066 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8067 if ($scfg->{path
}) {
8068 $name .= ".$dst_format";
8071 $size = PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
;
8072 } elsif ($dst_drivename eq 'efidisk0') {
8073 $size = $efisize or die "internal error - need to specify EFI disk size\n";
8074 } elsif ($dst_drivename eq 'tpmstate0') {
8075 $dst_format = 'raw';
8076 $size = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8078 clone_disk_check_io_uring
($drive, $storecfg, $src_storeid, $storeid, $use_drive_mirror);
8080 ($size) = PVE
::Storage
::volume_size_info
($storecfg, $drive->{file
}, 10);
8082 $newvolid = PVE
::Storage
::vdisk_alloc
(
8083 $storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024)
8085 push @$newvollist, $newvolid;
8087 PVE
::Storage
::activate_volumes
($storecfg, [$newvolid]);
8089 if (drive_is_cloudinit
($drive)) {
8090 # when cloning multiple disks (e.g. during clone_vm) it might be the last disk
8091 # if this is the case, we have to complete any block-jobs still there from
8092 # previous drive-mirrors
8093 if (($completion eq 'complete') && (scalar(keys %$jobs) > 0)) {
8094 qemu_drive_mirror_monitor
($vmid, $newvmid, $jobs, $completion, $qga);
8099 my $sparseinit = PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $newvolid);
8100 if ($use_drive_mirror) {
8101 qemu_drive_mirror
($vmid, $src_drivename, $newvolid, $newvmid, $sparseinit, $jobs,
8102 $completion, $qga, $bwlimit);
8104 if ($dst_drivename eq 'efidisk0') {
8105 # the relevant data on the efidisk may be smaller than the source
8106 # e.g. on RBD/ZFS, so we use dd to copy only the amount
8107 # that is given by the OVMF_VARS.fd
8108 my $src_path = PVE
::Storage
::path
($storecfg, $drive->{file
}, $snapname);
8109 my $dst_path = PVE
::Storage
::path
($storecfg, $newvolid);
8111 my $src_format = (PVE
::Storage
::parse_volname
($storecfg, $drive->{file
}))[6];
8113 # better for Ceph if block size is not too small, see bug #3324
8116 my $cmd = ['qemu-img', 'dd', '-n', '-O', $dst_format];
8118 if ($src_format eq 'qcow2' && $snapname) {
8119 die "cannot clone qcow2 EFI disk snapshot - requires QEMU >= 6.2\n"
8120 if !min_version
(kvm_user_version
(), 6, 2);
8121 push $cmd->@*, '-l', $snapname;
8123 push $cmd->@*, "bs=$bs", "osize=$size", "if=$src_path", "of=$dst_path";
8126 qemu_img_convert
($drive->{file
}, $newvolid, $size, $snapname, $sparseinit, $bwlimit);
8132 my ($size) = eval { PVE
::Storage
::volume_size_info
($storecfg, $newvolid, 10) };
8134 my $disk = dclone
($drive);
8135 delete $disk->{format
};
8136 $disk->{file
} = $newvolid;
8137 $disk->{size
} = $size if defined($size);
8142 sub get_running_qemu_version
{
8144 my $res = mon_cmd
($vmid, "query-version");
8145 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
8148 sub qemu_use_old_bios_files
{
8149 my ($machine_type) = @_;
8151 return if !$machine_type;
8153 my $use_old_bios_files = undef;
8155 if ($machine_type =~ m/^(\S+)\.pxe$/) {
8157 $use_old_bios_files = 1;
8159 my $version = extract_version
($machine_type, kvm_user_version
());
8160 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
8161 # load new efi bios files on migration. So this hack is required to allow
8162 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
8163 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
8164 $use_old_bios_files = !min_version
($version, 2, 4);
8167 return ($use_old_bios_files, $machine_type);
8170 sub get_efivars_size
{
8171 my ($conf, $efidisk) = @_;
8173 my $arch = get_vm_arch
($conf);
8174 $efidisk //= $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
8175 my $smm = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
8176 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8177 return -s
$ovmf_vars;
8180 sub update_efidisk_size
{
8183 return if !defined($conf->{efidisk0
});
8185 my $disk = PVE
::QemuServer
::parse_drive
('efidisk0', $conf->{efidisk0
});
8186 $disk->{size
} = get_efivars_size
($conf);
8187 $conf->{efidisk0
} = print_drive
($disk);
8192 sub update_tpmstate_size
{
8195 my $disk = PVE
::QemuServer
::parse_drive
('tpmstate0', $conf->{tpmstate0
});
8196 $disk->{size
} = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8197 $conf->{tpmstate0
} = print_drive
($disk);
8200 sub create_efidisk
($$$$$$$) {
8201 my ($storecfg, $storeid, $vmid, $fmt, $arch, $efidisk, $smm) = @_;
8203 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8205 my $vars_size_b = -s
$ovmf_vars;
8206 my $vars_size = PVE
::Tools
::convert_size
($vars_size_b, 'b' => 'kb');
8207 my $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
8208 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
8210 qemu_img_convert
($ovmf_vars, $volid, $vars_size_b, undef, 0);
8211 my ($size) = PVE
::Storage
::volume_size_info
($storecfg, $volid, 3);
8213 return ($volid, $size/1024);
8216 sub vm_iothreads_list
{
8219 my $res = mon_cmd
($vmid, 'query-iothreads');
8222 foreach my $iothread (@$res) {
8223 $iothreads->{ $iothread->{id
} } = $iothread->{"thread-id"};
8230 my ($conf, $drive) = @_;
8234 if (!$conf->{scsihw
} || ($conf->{scsihw
} =~ m/^lsi/)) {
8236 } elsif ($conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single')) {
8242 my $controller = int($drive->{index} / $maxdev);
8243 my $controller_prefix = ($conf->{scsihw
} && $conf->{scsihw
} eq 'virtio-scsi-single')
8247 return ($maxdev, $controller, $controller_prefix);
8250 sub resolve_dst_disk_format
{
8251 my ($storecfg, $storeid, $src_volname, $format) = @_;
8252 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
8255 # if no target format is specified, use the source disk format as hint
8257 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8258 $format = qemu_img_format
($scfg, $src_volname);
8264 # test if requested format is supported - else use default
8265 my $supported = grep { $_ eq $format } @$validFormats;
8266 $format = $defFormat if !$supported;
8270 # NOTE: if this logic changes, please update docs & possibly gui logic
8271 sub find_vmstate_storage
{
8272 my ($conf, $storecfg) = @_;
8274 # first, return storage from conf if set
8275 return $conf->{vmstatestorage
} if $conf->{vmstatestorage
};
8277 my ($target, $shared, $local);
8279 foreach_storage_used_by_vm
($conf, sub {
8281 my $scfg = PVE
::Storage
::storage_config
($storecfg, $sid);
8282 my $dst = $scfg->{shared
} ? \
$shared : \
$local;
8283 $$dst = $sid if !$$dst || $scfg->{path
}; # prefer file based storage
8286 # second, use shared storage where VM has at least one disk
8287 # third, use local storage where VM has at least one disk
8288 # fall back to local storage
8289 $target = $shared // $local // 'local';
8295 my ($uuid, $uuid_str);
8296 UUID
::generate
($uuid);
8297 UUID
::unparse
($uuid, $uuid_str);
8301 sub generate_smbios1_uuid
{
8302 return "uuid=".generate_uuid
();
8308 mon_cmd
($vmid, 'nbd-server-stop');
8311 sub create_reboot_request
{
8313 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
8314 or die "failed to create reboot trigger file: $!\n";
8318 sub clear_reboot_request
{
8320 my $path = "/run/qemu-server/$vmid.reboot";
8323 $res = unlink($path);
8324 die "could not remove reboot request for $vmid: $!"
8325 if !$res && $! != POSIX
::ENOENT
;
8330 sub bootorder_from_legacy
{
8331 my ($conf, $bootcfg) = @_;
8333 my $boot = $bootcfg->{legacy
} || $boot_fmt->{legacy
}->{default};
8334 my $bootindex_hash = {};
8336 foreach my $o (split(//, $boot)) {
8337 $bootindex_hash->{$o} = $i*100;
8343 PVE
::QemuConfig-
>foreach_volume($conf, sub {
8344 my ($ds, $drive) = @_;
8346 if (drive_is_cdrom
($drive, 1)) {
8347 if ($bootindex_hash->{d
}) {
8348 $bootorder->{$ds} = $bootindex_hash->{d
};
8349 $bootindex_hash->{d
} += 1;
8351 } elsif ($bootindex_hash->{c
}) {
8352 $bootorder->{$ds} = $bootindex_hash->{c
}
8353 if $conf->{bootdisk
} && $conf->{bootdisk
} eq $ds;
8354 $bootindex_hash->{c
} += 1;
8358 if ($bootindex_hash->{n
}) {
8359 for (my $i = 0; $i < $MAX_NETS; $i++) {
8360 my $netname = "net$i";
8361 next if !$conf->{$netname};
8362 $bootorder->{$netname} = $bootindex_hash->{n
};
8363 $bootindex_hash->{n
} += 1;
8370 # Generate default device list for 'boot: order=' property. Matches legacy
8371 # default boot order, but with explicit device names. This is important, since
8372 # the fallback for when neither 'order' nor the old format is specified relies
8373 # on 'bootorder_from_legacy' above, and it would be confusing if this diverges.
8374 sub get_default_bootdevices
{
8380 my $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 0);
8381 push @ret, $first if $first;
8384 $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 1);
8385 push @ret, $first if $first;
8388 for (my $i = 0; $i < $MAX_NETS; $i++) {
8389 my $netname = "net$i";
8390 next if !$conf->{$netname};
8391 push @ret, $netname;
8398 sub device_bootorder
{
8401 return bootorder_from_legacy
($conf) if !defined($conf->{boot
});
8403 my $boot = parse_property_string
($boot_fmt, $conf->{boot
});
8406 if (!defined($boot) || $boot->{legacy
}) {
8407 $bootorder = bootorder_from_legacy
($conf, $boot);
8408 } elsif ($boot->{order
}) {
8409 my $i = 100; # start at 100 to allow user to insert devices before us with -args
8410 for my $dev (PVE
::Tools
::split_list
($boot->{order
})) {
8411 $bootorder->{$dev} = $i++;
8418 sub register_qmeventd_handle
{
8422 my $peer = "/var/run/qmeventd.sock";
8427 $fh = IO
::Socket
::UNIX-
>new(Peer
=> $peer, Blocking
=> 0, Timeout
=> 1);
8429 if ($! != EINTR
&& $! != EAGAIN
) {
8430 die "unable to connect to qmeventd socket (vmid: $vmid) - $!\n";
8433 die "unable to connect to qmeventd socket (vmid: $vmid) - timeout "
8434 . "after $count retries\n";
8439 # send handshake to mark VM as backing up
8440 print $fh to_json
({vzdump
=> {vmid
=> "$vmid"}});
8442 # return handle to be closed later when inhibit is no longer required
8446 # bash completion helper
8448 sub complete_backup_archives
{
8449 my ($cmdname, $pname, $cvalue) = @_;
8451 my $cfg = PVE
::Storage
::config
();
8455 if ($cvalue =~ m/^([^:]+):/) {
8459 my $data = PVE
::Storage
::template_list
($cfg, $storeid, 'backup');
8462 foreach my $id (keys %$data) {
8463 foreach my $item (@{$data->{$id}}) {
8464 next if $item->{format
} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
8465 push @$res, $item->{volid
} if defined($item->{volid
});
8472 my $complete_vmid_full = sub {
8475 my $idlist = vmstatus
();
8479 foreach my $id (keys %$idlist) {
8480 my $d = $idlist->{$id};
8481 if (defined($running)) {
8482 next if $d->{template
};
8483 next if $running && $d->{status
} ne 'running';
8484 next if !$running && $d->{status
} eq 'running';
8493 return &$complete_vmid_full();
8496 sub complete_vmid_stopped
{
8497 return &$complete_vmid_full(0);
8500 sub complete_vmid_running
{
8501 return &$complete_vmid_full(1);
8504 sub complete_storage
{
8506 my $cfg = PVE
::Storage
::config
();
8507 my $ids = $cfg->{ids
};
8510 foreach my $sid (keys %$ids) {
8511 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, undef, 1);
8512 next if !$ids->{$sid}->{content
}->{images
};
8519 sub complete_migration_storage
{
8520 my ($cmd, $param, $current_value, $all_args) = @_;
8522 my $targetnode = @$all_args[1];
8524 my $cfg = PVE
::Storage
::config
();
8525 my $ids = $cfg->{ids
};
8528 foreach my $sid (keys %$ids) {
8529 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, $targetnode, 1);
8530 next if !$ids->{$sid}->{content
}->{images
};
8539 my $qmpstatus = eval {
8540 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid);
8541 mon_cmd
($vmid, "query-status");
8544 return $qmpstatus && $qmpstatus->{status
} eq "paused";
8547 sub check_volume_storage_type
{
8548 my ($storecfg, $vol) = @_;
8550 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($vol);
8551 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8552 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $vol);
8554 die "storage '$storeid' does not support content-type '$vtype'\n"
8555 if !$scfg->{content
}->{$vtype};
8560 sub add_nets_bridge_fdb
{
8561 my ($conf, $vmid) = @_;
8563 for my $opt (keys %$conf) {
8564 next if $opt !~ m/^net(\d+)$/;
8565 my $iface = "tap${vmid}i$1";
8566 # NOTE: expect setups with learning off to *not* use auto-random-generation of MAC on start
8567 my $net = parse_net
($conf->{$opt}, 1) or next;
8569 my $mac = $net->{macaddr
};
8571 log_warn
("MAC learning disabled, but vNIC '$iface' has no static MAC to add to forwarding DB!")
8572 if !file_read_firstline
("/sys/class/net/$iface/brport/learning");
8576 my $bridge = $net->{bridge
};
8578 PVE
::Network
::SDN
::Zones
::add_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8579 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8580 PVE
::Network
::add_bridge_fdb
($iface, $mac, $net->{firewall
});
8585 sub del_nets_bridge_fdb
{
8586 my ($conf, $vmid) = @_;
8588 for my $opt (keys %$conf) {
8589 next if $opt !~ m/^net(\d+)$/;
8590 my $iface = "tap${vmid}i$1";
8592 my $net = parse_net
($conf->{$opt}) or next;
8593 my $mac = $net->{macaddr
} or next;
8595 my $bridge = $net->{bridge
};
8597 PVE
::Network
::SDN
::Zones
::del_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8598 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8599 PVE
::Network
::del_bridge_fdb
($iface, $mac, $net->{firewall
});