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.",
166 description
=> "Select the agent type",
170 enum
=> [qw(virtio isa)],
176 description
=> "Select the VGA type.",
181 enum
=> [qw(cirrus qxl qxl2 qxl3 qxl4 none serial0 serial1 serial2 serial3 std virtio virtio-gl vmware)],
184 description
=> "Sets the VGA memory (in MiB). Has no effect with serial display.",
196 description
=> "The size of the file in MB.",
200 pattern
=> '[a-zA-Z0-9\-]+',
202 format_description
=> 'string',
203 description
=> "The name of the file. Will be prefixed with 'pve-shm-'. Default is the VMID. Will be deleted when the VM is stopped.",
210 enum
=> [qw(ich9-intel-hda intel-hda AC97)],
211 description
=> "Configure an audio device."
215 enum
=> ['spice', 'none'],
218 description
=> "Driver backend for the audio device."
222 my $spice_enhancements_fmt = {
227 description
=> "Enable folder sharing via SPICE. Needs Spice-WebDAV daemon installed in the VM."
231 enum
=> ['off', 'all', 'filter'],
234 description
=> "Enable video streaming. Uses compression for detected video streams."
241 enum
=> ['/dev/urandom', '/dev/random', '/dev/hwrng'],
243 description
=> "The file on the host to gather entropy from. In most cases '/dev/urandom'"
244 ." should be preferred over '/dev/random' to avoid entropy-starvation issues on the"
245 ." host. Using urandom does *not* decrease security in any meaningful way, as it's"
246 ." still seeded from real entropy, and the bytes provided will most likely be mixed"
247 ." with real entropy on the guest as well. '/dev/hwrng' can be used to pass through"
248 ." a hardware RNG from the host.",
252 description
=> "Maximum bytes of entropy allowed to get injected into the guest every"
253 ." 'period' milliseconds. Prefer a lower value when using '/dev/random' as source. Use"
254 ." `0` to disable limiting (potentially dangerous!).",
257 # default is 1 KiB/s, provides enough entropy to the guest to avoid boot-starvation issues
258 # (e.g. systemd etc...) while allowing no chance of overwhelming the host, provided we're
259 # reading from /dev/urandom
264 description
=> "Every 'period' milliseconds the entropy-injection quota is reset, allowing"
265 ." the guest to retrieve another 'max_bytes' of entropy.",
271 my $meta_info_fmt = {
274 description
=> "The guest creation timestamp as UNIX epoch time",
280 description
=> "The QEMU (machine) version from the time this VM was created.",
281 pattern
=> '\d+(\.\d+)+',
290 description
=> "Specifies whether a VM will be started during system bootup.",
296 description
=> "Automatic restart after crash (currently ignored).",
301 type
=> 'string', format
=> 'pve-hotplug-features',
302 description
=> "Selectively enable hotplug features. This is a comma separated list of"
303 ." hotplug features: 'network', 'disk', 'cpu', 'memory', 'usb' and 'cloudinit'. Use '0' to disable"
304 ." hotplug completely. Using '1' as value is an alias for the default `network,disk,usb`."
305 ." USB hotplugging is possible for guests with machine version >= 7.1 and ostype l26 or"
307 default => 'network,disk,usb',
312 description
=> "Allow reboot. If set to '0' the VM exit on reboot.",
318 description
=> "Lock/unlock the VM.",
319 enum
=> [qw(backup clone create migrate rollback snapshot snapshot-delete suspending suspended)],
324 description
=> "Limit of CPU usage.",
325 verbose_description
=> "Limit of CPU usage.\n\nNOTE: If the computer has 2 CPUs, it has"
326 ." total of '2' CPU time. Value '0' indicates no CPU limit.",
334 description
=> "CPU weight for a VM, will be clamped to [1, 10000] in cgroup v2.",
335 verbose_description
=> "CPU weight for a VM. Argument is used in the kernel fair scheduler."
336 ." The larger the number is, the more CPU time this VM gets. Number is relative to"
337 ." weights of all the other running VMs.",
340 default => 'cgroup v1: 1024, cgroup v2: 100',
345 description
=> "Amount of RAM for the VM in MiB. This is the maximum available memory when"
346 ." you use the balloon device.",
353 description
=> "Amount of target RAM for the VM in MiB. Using zero disables the ballon driver.",
359 description
=> "Amount of memory shares for auto-ballooning. The larger the number is, the"
360 ." more memory this VM gets. Number is relative to weights of all other running VMs."
361 ." Using zero disables auto-ballooning. Auto-ballooning is done by pvestatd.",
369 description
=> "Keyboard layout for VNC server. This option is generally not required and"
370 ." is often better handled from within the guest OS.",
371 enum
=> PVE
::Tools
::kvmkeymaplist
(),
376 type
=> 'string', format
=> 'dns-name',
377 description
=> "Set a name for the VM. Only used on the configuration web interface.",
382 description
=> "SCSI controller model",
383 enum
=> [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
389 description
=> "Description for the VM. Shown in the web-interface VM's summary."
390 ." This is saved as comment inside the configuration file.",
391 maxLength
=> 1024 * 8,
396 enum
=> [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 win11 l24 l26 solaris)],
397 description
=> "Specify guest operating system.",
398 verbose_description
=> <<EODESC,
399 Specify guest operating system. This is used to enable special
400 optimization/features for specific operating systems:
403 other;; unspecified OS
404 wxp;; Microsoft Windows XP
405 w2k;; Microsoft Windows 2000
406 w2k3;; Microsoft Windows 2003
407 w2k8;; Microsoft Windows 2008
408 wvista;; Microsoft Windows Vista
409 win7;; Microsoft Windows 7
410 win8;; Microsoft Windows 8/2012/2012r2
411 win10;; Microsoft Windows 10/2016/2019
412 win11;; Microsoft Windows 11/2022
413 l24;; Linux 2.4 Kernel
414 l26;; Linux 2.6 - 6.X Kernel
415 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
420 type
=> 'string', format
=> 'pve-qm-boot',
421 description
=> "Specify guest boot order. Use the 'order=' sub-property as usage with no"
422 ." key or 'legacy=' is deprecated.",
426 type
=> 'string', format
=> 'pve-qm-bootdisk',
427 description
=> "Enable booting from specified disk. Deprecated: Use 'boot: order=foo;bar' instead.",
428 pattern
=> '(ide|sata|scsi|virtio)\d+',
433 description
=> "The number of CPUs. Please use option -sockets instead.",
440 description
=> "The number of CPU sockets.",
447 description
=> "The number of cores per socket.",
454 description
=> "Enable/disable NUMA.",
460 description
=> "Enable/disable hugepages memory.",
461 enum
=> [qw(any 2 1024)],
467 description
=> "Use together with hugepages. If enabled, hugepages will not not be deleted"
468 ." after VM shutdown and can be used for subsequent starts.",
473 description
=> "Number of hotplugged vcpus.",
480 description
=> "Enable/disable ACPI.",
485 description
=> "Enable/disable communication with the QEMU Guest Agent and its properties.",
487 format
=> $agent_fmt,
492 description
=> "Enable/disable KVM hardware virtualization.",
498 description
=> "Enable/disable time drift fix.",
504 description
=> "Set the real time clock (RTC) to local time. This is enabled by default if"
505 ." the `ostype` indicates a Microsoft Windows OS.",
510 description
=> "Freeze CPU at startup (use 'c' monitor command to start execution).",
514 type
=> 'string', format
=> $vga_fmt,
515 description
=> "Configure the VGA hardware.",
516 verbose_description
=> "Configure the VGA Hardware. If you want to use high resolution"
517 ." modes (>= 1280x1024x16) you may need to increase the vga memory option. Since QEMU"
518 ." 2.9 the default VGA display type is 'std' for all OS types besides some Windows"
519 ." versions (XP and older) which use 'cirrus'. The 'qxl' option enables the SPICE"
520 ." display server. For win* OS you can select how many independent displays you want,"
521 ." Linux guests can add displays them self.\nYou can also run without any graphic card,"
522 ." using a serial device as terminal.",
526 type
=> 'string', format
=> 'pve-qm-watchdog',
527 description
=> "Create a virtual hardware watchdog device.",
528 verbose_description
=> "Create a virtual hardware watchdog device. Once enabled (by a guest"
529 ." action), the watchdog must be periodically polled by an agent inside the guest or"
530 ." else the watchdog will reset the guest (or execute the respective action specified)",
535 typetext
=> "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
536 description
=> "Set the initial date of the real time clock. Valid format for date are:"
537 ."'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
538 pattern
=> '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
541 startup
=> get_standard_option
('pve-startup-order'),
545 description
=> "Enable/disable Template.",
551 description
=> "Arbitrary arguments passed to kvm.",
552 verbose_description
=> <<EODESCR,
553 Arbitrary arguments passed to kvm, for example:
555 args: -no-reboot -no-hpet
557 NOTE: this option is for experts only.
564 description
=> "Enable/disable the USB tablet device.",
565 verbose_description
=> "Enable/disable the USB tablet device. This device is usually needed"
566 ." to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with"
567 ." normal VNC clients. If you're running lots of console-only guests on one host, you"
568 ." may consider disabling this to save some context switches. This is turned off by"
569 ." default if you use spice (`qm set <vmid> --vga qxl`).",
574 description
=> "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
578 migrate_downtime
=> {
581 description
=> "Set maximum tolerated downtime (in seconds) for migrations.",
587 type
=> 'string', format
=> 'pve-qm-ide',
588 typetext
=> '<volume>',
589 description
=> "This is an alias for option -ide2",
593 description
=> "Emulated CPU type.",
595 format
=> 'pve-vm-cpu-conf',
597 parent
=> get_standard_option
('pve-snapshot-name', {
599 description
=> "Parent snapshot name. This is used internally, and should not be modified.",
603 description
=> "Timestamp for snapshots.",
609 type
=> 'string', format
=> 'pve-volume-id',
610 description
=> "Reference to a volume which stores the VM state. This is used internally"
613 vmstatestorage
=> get_standard_option
('pve-storage-id', {
614 description
=> "Default storage for VM state volumes/files.",
617 runningmachine
=> get_standard_option
('pve-qemu-machine', {
618 description
=> "Specifies the QEMU machine type of the running vm. This is used internally"
622 description
=> "Specifies the QEMU '-cpu' parameter of the running vm. This is used"
623 ." internally for snapshots.",
626 pattern
=> $PVE::QemuServer
::CPUConfig
::qemu_cmdline_cpu_re
,
627 format_description
=> 'QEMU -cpu parameter'
629 machine
=> get_standard_option
('pve-qemu-machine'),
631 description
=> "Virtual processor architecture. Defaults to the host.",
634 enum
=> [qw(x86_64 aarch64)],
637 description
=> "Specify SMBIOS type 1 fields.",
638 type
=> 'string', format
=> 'pve-qm-smbios1',
645 description
=> "Sets the protection flag of the VM. This will disable the remove VM and"
646 ." remove disk operations.",
652 enum
=> [ qw(seabios ovmf) ],
653 description
=> "Select BIOS implementation.",
654 default => 'seabios',
658 pattern
=> '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
659 format_description
=> 'UUID',
660 description
=> "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0'"
661 ." to disable explicitly.",
662 verbose_description
=> "The VM generation ID (vmgenid) device exposes a 128-bit integer"
663 ." value identifier to the guest OS. This allows to notify the guest operating system"
664 ." when the virtual machine is executed with a different configuration (e.g. snapshot"
665 ." execution or creation from a template). The guest operating system notices the"
666 ." change, and is then able to react as appropriate by marking its copies of"
667 ." distributed databases as dirty, re-initializing its random number generator, etc.\n"
668 ."Note that auto-creation only works when done through API/CLI create or update methods"
669 .", but not when manually editing the config file.",
670 default => "1 (autogenerated)",
675 format
=> 'pve-volume-id',
677 description
=> "Script that will be executed during various steps in the vms lifetime.",
681 format
=> $ivshmem_fmt,
682 description
=> "Inter-VM shared memory. Useful for direct communication between VMs, or to"
688 format
=> $audio_fmt,
689 description
=> "Configure a audio device, useful in combination with QXL/Spice.",
692 spice_enhancements
=> {
694 format
=> $spice_enhancements_fmt,
695 description
=> "Configure additional enhancements for SPICE.",
699 type
=> 'string', format
=> 'pve-tag-list',
700 description
=> 'Tags of the VM. This is only meta information.',
706 description
=> "Configure a VirtIO-based Random Number Generator.",
711 format
=> $meta_info_fmt,
712 description
=> "Some (read-only) meta-information about this guest.",
716 type
=> 'string', format
=> 'pve-cpuset',
717 description
=> "List of host cores used to execute guest processes, for example: 0,5,8-11",
726 description
=> 'Specify a custom file containing all meta data passed to the VM via"
727 ." cloud-init. This is provider specific meaning configdrive2 and nocloud differ.',
728 format
=> 'pve-volume-id',
729 format_description
=> 'volume',
734 description
=> 'To pass a custom file containing all network data to the VM via cloud-init.',
735 format
=> 'pve-volume-id',
736 format_description
=> 'volume',
741 description
=> 'To pass a custom file containing all user data to the VM via cloud-init.',
742 format
=> 'pve-volume-id',
743 format_description
=> 'volume',
748 description
=> 'To pass a custom file containing all vendor data to the VM via cloud-init.',
749 format
=> 'pve-volume-id',
750 format_description
=> 'volume',
753 PVE
::JSONSchema
::register_format
('pve-qm-cicustom', $cicustom_fmt);
755 my $confdesc_cloudinit = {
759 description
=> 'Specifies the cloud-init configuration format. The default depends on the'
760 .' configured operating system type (`ostype`. We use the `nocloud` format for Linux,'
761 .' and `configdrive2` for windows.',
762 enum
=> ['configdrive2', 'nocloud', 'opennebula'],
767 description
=> "cloud-init: User name to change ssh keys and password for instead of the"
768 ." image's configured default user.",
773 description
=> 'cloud-init: Password to assign the user. Using this is generally not'
774 .' recommended. Use ssh keys instead. Also note that older cloud-init versions do not'
775 .' support hashed passwords.',
780 description
=> 'cloud-init: Specify custom files to replace the automatically generated'
782 format
=> 'pve-qm-cicustom',
787 description
=> 'cloud-init: Sets DNS search domains for a container. Create will'
788 .' automatically use the setting from the host if neither searchdomain nor nameserver'
793 type
=> 'string', format
=> 'address-list',
794 description
=> 'cloud-init: Sets DNS server IP address for a container. Create will'
795 .' automatically use the setting from the host if neither searchdomain nor nameserver'
801 format
=> 'urlencoded',
802 description
=> "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
806 # what about other qemu settings ?
808 #machine => 'string',
821 ##soundhw => 'string',
823 while (my ($k, $v) = each %$confdesc) {
824 PVE
::JSONSchema
::register_standard_option
("pve-qm-$k", $v);
827 my $MAX_USB_DEVICES = 14;
829 my $MAX_SERIAL_PORTS = 4;
830 my $MAX_PARALLEL_PORTS = 3;
836 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
837 description
=> "CPUs accessing this NUMA node.",
838 format_description
=> "id[-id];...",
842 description
=> "Amount of memory this NUMA node provides.",
847 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
848 description
=> "Host NUMA nodes to use.",
849 format_description
=> "id[-id];...",
854 enum
=> [qw(preferred bind interleave)],
855 description
=> "NUMA allocation policy.",
859 PVE
::JSONSchema
::register_format
('pve-qm-numanode', $numa_fmt);
862 type
=> 'string', format
=> $numa_fmt,
863 description
=> "NUMA topology.",
865 PVE
::JSONSchema
::register_standard_option
("pve-qm-numanode", $numadesc);
867 for (my $i = 0; $i < $MAX_NUMA; $i++) {
868 $confdesc->{"numa$i"} = $numadesc;
871 my $nic_model_list = [
887 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
889 my $net_fmt_bridge_descr = <<__EOD__;
890 Bridge to attach the network device to. The Proxmox VE standard bridge
893 If you do not specify a bridge, we create a kvm user (NATed) network
894 device, which provides DHCP and DNS services. The following addresses
901 The DHCP server assign addresses to the guest starting from 10.0.2.15.
905 macaddr
=> get_standard_option
('mac-addr', {
906 description
=> "MAC address. That address must be unique withing your network. This is"
907 ." automatically generated if not specified.",
911 description
=> "Network Card Model. The 'virtio' model provides the best performance with"
912 ." very low CPU overhead. If your guest does not support this driver, it is usually"
913 ." best to use 'e1000'.",
914 enum
=> $nic_model_list,
917 (map { $_ => { keyAlias
=> 'model', alias
=> 'macaddr' }} @$nic_model_list),
918 bridge
=> get_standard_option
('pve-bridge-id', {
919 description
=> $net_fmt_bridge_descr,
924 minimum
=> 0, maximum
=> 64,
925 description
=> 'Number of packet queues to be used on the device.',
931 description
=> "Rate limit in mbps (megabytes per second) as floating point number.",
936 minimum
=> 1, maximum
=> 4094,
937 description
=> 'VLAN tag to apply to packets on this interface.',
942 pattern
=> qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
943 description
=> 'VLAN trunks to pass through this interface.',
944 format_description
=> 'vlanid[;vlanid...]',
949 description
=> 'Whether this interface should be protected by the firewall.',
954 description
=> 'Whether this interface should be disconnected (like pulling the plug).',
959 minimum
=> 1, maximum
=> 65520,
960 description
=> "Force MTU, for VirtIO only. Set to '1' to use the bridge MTU",
967 type
=> 'string', format
=> $net_fmt,
968 description
=> "Specify network devices.",
971 PVE
::JSONSchema
::register_standard_option
("pve-qm-net", $netdesc);
976 format
=> 'pve-ipv4-config',
977 format_description
=> 'IPv4Format/CIDR',
978 description
=> 'IPv4 address in CIDR format.',
985 format_description
=> 'GatewayIPv4',
986 description
=> 'Default gateway for IPv4 traffic.',
992 format
=> 'pve-ipv6-config',
993 format_description
=> 'IPv6Format/CIDR',
994 description
=> 'IPv6 address in CIDR format.',
1001 format_description
=> 'GatewayIPv6',
1002 description
=> 'Default gateway for IPv6 traffic.',
1007 PVE
::JSONSchema
::register_format
('pve-qm-ipconfig', $ipconfig_fmt);
1008 my $ipconfigdesc = {
1010 type
=> 'string', format
=> 'pve-qm-ipconfig',
1011 description
=> <<'EODESCR',
1012 cloud-init: Specify IP addresses and gateways for the corresponding interface.
1014 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
1016 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit
1017 gateway should be provided.
1018 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration. This requires
1019 cloud-init 19.4 or newer.
1021 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using
1025 PVE
::JSONSchema
::register_standard_option
("pve-qm-ipconfig", $netdesc);
1027 for (my $i = 0; $i < $MAX_NETS; $i++) {
1028 $confdesc->{"net$i"} = $netdesc;
1029 $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
1032 foreach my $key (keys %$confdesc_cloudinit) {
1033 $confdesc->{$key} = $confdesc_cloudinit->{$key};
1036 PVE
::JSONSchema
::register_format
('pve-cpuset', \
&pve_verify_cpuset
);
1037 sub pve_verify_cpuset
{
1038 my ($set_text, $noerr) = @_;
1040 my ($count, $members) = eval { PVE
::CpuSet
::parse_cpuset
($set_text) };
1044 die "unable to parse cpuset option\n";
1047 return PVE
::CpuSet-
>new($members)->short_string();
1050 PVE
::JSONSchema
::register_format
('pve-volume-id-or-qm-path', \
&verify_volume_id_or_qm_path
);
1051 sub verify_volume_id_or_qm_path
{
1052 my ($volid, $noerr) = @_;
1054 return $volid if $volid eq 'none' || $volid eq 'cdrom';
1056 return verify_volume_id_or_absolute_path
($volid, $noerr);
1059 PVE
::JSONSchema
::register_format
('pve-volume-id-or-absolute-path', \
&verify_volume_id_or_absolute_path
);
1060 sub verify_volume_id_or_absolute_path
{
1061 my ($volid, $noerr) = @_;
1063 return $volid if $volid =~ m
|^/|;
1065 $volid = eval { PVE
::JSONSchema
::check_format
('pve-volume-id', $volid, '') };
1076 type
=> 'string', format
=> 'pve-qm-usb-device',
1077 format_description
=> 'HOSTUSBDEVICE|spice',
1078 description
=> <<EODESCR,
1079 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1081 'bus-port(.port)*' (decimal numbers) or
1082 'vendor_id:product_id' (hexadeciaml numbers) or
1085 You can use the 'lsusb -t' command to list existing usb devices.
1087 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1088 machines - use with special care.
1090 The value 'spice' can be used to add a usb redirection devices for spice.
1096 description
=> "Specifies whether if given host option is a USB3 device or port."
1097 ." For modern guests (machine version >= 7.1 and ostype l26 and windows > 7), this flag"
1098 ." is irrelevant (all devices are plugged into a xhci controller).",
1105 type
=> 'string', format
=> $usb_fmt,
1106 description
=> "Configure an USB device (n is 0 to 4, for machine version >= 7.1 and ostype"
1107 ." l26 or windows > 7, n can be up to 14).",
1109 PVE
::JSONSchema
::register_standard_option
("pve-qm-usb", $usbdesc);
1114 pattern
=> '(/dev/.+|socket)',
1115 description
=> "Create a serial device inside the VM (n is 0 to 3)",
1116 verbose_description
=> <<EODESCR,
1117 Create a serial device inside the VM (n is 0 to 3), and pass through a
1118 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1119 host side (use 'qm terminal' to open a terminal connection).
1121 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines -
1122 use with special care.
1124 CAUTION: Experimental! User reported problems with this option.
1131 pattern
=> '/dev/parport\d+|/dev/usb/lp\d+',
1132 description
=> "Map host parallel devices (n is 0 to 2).",
1133 verbose_description
=> <<EODESCR,
1134 Map host parallel devices (n is 0 to 2).
1136 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1137 machines - use with special care.
1139 CAUTION: Experimental! User reported problems with this option.
1143 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
1144 $confdesc->{"parallel$i"} = $paralleldesc;
1147 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
1148 $confdesc->{"serial$i"} = $serialdesc;
1151 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
1152 $confdesc->{"hostpci$i"} = $PVE::QemuServer
::PCI
::hostpcidesc
;
1155 for my $key (keys %{$PVE::QemuServer
::Drive
::drivedesc_hash
}) {
1156 $confdesc->{$key} = $PVE::QemuServer
::Drive
::drivedesc_hash-
>{$key};
1159 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
1160 $confdesc->{"usb$i"} = $usbdesc;
1168 description
=> "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n)."
1169 . " Deprecated, use 'order=' instead.",
1170 pattern
=> '[acdn]{1,4}',
1171 format_description
=> "[acdn]{1,4}",
1173 # note: this is also the fallback if boot: is not given at all
1179 format
=> 'pve-qm-bootdev-list',
1180 format_description
=> "device[;device...]",
1181 description
=> <<EODESC,
1182 The guest will attempt to boot from devices in the order they appear here.
1184 Disks, optical drives and passed-through storage USB devices will be directly
1185 booted from, NICs will load PXE, and PCIe devices will either behave like disks
1186 (e.g. NVMe) or load an option ROM (e.g. RAID controller, hardware NIC).
1188 Note that only devices in this list will be marked as bootable and thus loaded
1189 by the guest firmware (BIOS/UEFI). If you require multiple disks for booting
1190 (e.g. software-raid), you need to specify all of them here.
1192 Overrides the deprecated 'legacy=[acdn]*' value when given.
1196 PVE
::JSONSchema
::register_format
('pve-qm-boot', $boot_fmt);
1198 PVE
::JSONSchema
::register_format
('pve-qm-bootdev', \
&verify_bootdev
);
1199 sub verify_bootdev
{
1200 my ($dev, $noerr) = @_;
1202 my $special = $dev =~ m/^efidisk/ || $dev =~ m/^tpmstate/;
1203 return $dev if PVE
::QemuServer
::Drive
::is_valid_drivename
($dev) && !$special;
1207 return 0 if $dev !~ m/^$base\d+$/;
1208 return 0 if !$confdesc->{$dev};
1212 return $dev if $check->("net");
1213 return $dev if $check->("usb");
1214 return $dev if $check->("hostpci");
1217 die "invalid boot device '$dev'\n";
1220 sub print_bootorder
{
1222 return "" if !@$devs;
1223 my $data = { order
=> join(';', @$devs) };
1224 return PVE
::JSONSchema
::print_property_string
($data, $boot_fmt);
1227 my $kvm_api_version = 0;
1230 return $kvm_api_version if $kvm_api_version;
1232 open my $fh, '<', '/dev/kvm' or return;
1234 # 0xae00 => KVM_GET_API_VERSION
1235 $kvm_api_version = ioctl($fh, 0xae00, 0);
1238 return $kvm_api_version;
1241 my $kvm_user_version = {};
1244 sub kvm_user_version
{
1247 $binary //= get_command_for_arch
(get_host_arch
()); # get the native arch by default
1248 my $st = stat($binary);
1250 my $cachedmtime = $kvm_mtime->{$binary} // -1;
1251 return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1252 $cachedmtime == $st->mtime;
1254 $kvm_user_version->{$binary} = 'unknown';
1255 $kvm_mtime->{$binary} = $st->mtime;
1259 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1260 $kvm_user_version->{$binary} = $2;
1264 eval { run_command
([$binary, '--version'], outfunc
=> $code); };
1267 return $kvm_user_version->{$binary};
1270 my sub extract_version
{
1271 my ($machine_type, $version) = @_;
1272 $version = kvm_user_version
() if !defined($version);
1273 return PVE
::QemuServer
::Machine
::extract_version
($machine_type, $version)
1276 sub kernel_has_vhost_net
{
1277 return -c
'/dev/vhost-net';
1282 return defined($confdesc->{$key});
1286 sub get_cdrom_path
{
1288 return $cdrom_path if defined($cdrom_path);
1290 $cdrom_path = first
{ -l
$_ } map { "/dev/cdrom$_" } ('', '1', '2');
1292 if (!defined($cdrom_path)) {
1293 log_warn
("no physical CD-ROM available, ignoring");
1301 my ($storecfg, $vmid, $cdrom) = @_;
1303 if ($cdrom eq 'cdrom') {
1304 return get_cdrom_path
();
1305 } elsif ($cdrom eq 'none') {
1307 } elsif ($cdrom =~ m
|^/|) {
1310 return PVE
::Storage
::path
($storecfg, $cdrom);
1314 # try to convert old style file names to volume IDs
1315 sub filename_to_volume_id
{
1316 my ($vmid, $file, $media) = @_;
1318 if (!($file eq 'none' || $file eq 'cdrom' ||
1319 $file =~ m
|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1321 return if $file =~ m
|/|;
1323 if ($media && $media eq 'cdrom') {
1324 $file = "local:iso/$file";
1326 $file = "local:$vmid/$file";
1333 sub verify_media_type
{
1334 my ($opt, $vtype, $media) = @_;
1339 if ($media eq 'disk') {
1341 } elsif ($media eq 'cdrom') {
1344 die "internal error";
1347 return if ($vtype eq $etype);
1349 raise_param_exc
({ $opt => "unexpected media type ($vtype != $etype)" });
1352 sub cleanup_drive_path
{
1353 my ($opt, $storecfg, $drive) = @_;
1355 # try to convert filesystem paths to volume IDs
1357 if (($drive->{file
} !~ m/^(cdrom|none)$/) &&
1358 ($drive->{file
} !~ m
|^/dev/.+|) &&
1359 ($drive->{file
} !~ m/^([^:]+):(.+)$/) &&
1360 ($drive->{file
} !~ m/^\d+$/)) {
1361 my ($vtype, $volid) = PVE
::Storage
::path_to_volume_id
($storecfg, $drive->{file
});
1362 raise_param_exc
({ $opt => "unable to associate path '$drive->{file}' to any storage"})
1364 $drive->{media
} = 'cdrom' if !$drive->{media
} && $vtype eq 'iso';
1365 verify_media_type
($opt, $vtype, $drive->{media
});
1366 $drive->{file
} = $volid;
1369 $drive->{media
} = 'cdrom' if !$drive->{media
} && $drive->{file
} =~ m/^(cdrom|none)$/;
1372 sub parse_hotplug_features
{
1377 return $res if $data eq '0';
1379 $data = $confdesc->{hotplug
}->{default} if $data eq '1';
1381 foreach my $feature (PVE
::Tools
::split_list
($data)) {
1382 if ($feature =~ m/^(network|disk|cpu|memory|usb|cloudinit)$/) {
1385 die "invalid hotplug feature '$feature'\n";
1391 PVE
::JSONSchema
::register_format
('pve-hotplug-features', \
&pve_verify_hotplug_features
);
1392 sub pve_verify_hotplug_features
{
1393 my ($value, $noerr) = @_;
1395 return $value if parse_hotplug_features
($value);
1399 die "unable to parse hotplug option\n";
1403 my($fh, $noerr) = @_;
1406 my $SG_GET_VERSION_NUM = 0x2282;
1408 my $versionbuf = "\x00" x
8;
1409 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1411 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1414 my $version = unpack("I", $versionbuf);
1415 if ($version < 30000) {
1416 die "scsi generic interface too old\n" if !$noerr;
1420 my $buf = "\x00" x
36;
1421 my $sensebuf = "\x00" x
8;
1422 my $cmd = pack("C x3 C x1", 0x12, 36);
1424 # see /usr/include/scsi/sg.h
1425 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";
1428 $sg_io_hdr_t, ord('S'), -3, length($cmd), length($sensebuf), 0, length($buf), $buf, $cmd, $sensebuf, 6000
1431 $ret = ioctl($fh, $SG_IO, $packet);
1433 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1437 my @res = unpack($sg_io_hdr_t, $packet);
1438 if ($res[17] || $res[18]) {
1439 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1444 $res->@{qw(type removable vendor product revision)} = unpack("C C x6 A8 A16 A4", $buf);
1446 $res->{removable
} = $res->{removable
} & 128 ?
1 : 0;
1447 $res->{type
} &= 0x1F;
1455 my $fh = IO
::File-
>new("+<$path") || return;
1456 my $res = scsi_inquiry
($fh, 1);
1462 sub print_tabletdevice_full
{
1463 my ($conf, $arch) = @_;
1465 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1467 # we use uhci for old VMs because tablet driver was buggy in older qemu
1469 if ($q35 || $arch eq 'aarch64') {
1475 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1478 sub print_keyboarddevice_full
{
1479 my ($conf, $arch) = @_;
1481 return if $arch ne 'aarch64';
1483 return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1486 my sub get_drive_id
{
1488 return "$drive->{interface}$drive->{index}";
1491 sub print_drivedevice_full
{
1492 my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1497 my $drive_id = get_drive_id
($drive);
1498 if ($drive->{interface
} eq 'virtio') {
1499 my $pciaddr = print_pci_addr
("$drive_id", $bridges, $arch, $machine_type);
1500 $device = "virtio-blk-pci,drive=drive-$drive_id,id=${drive_id}${pciaddr}";
1501 $device .= ",iothread=iothread-$drive_id" if $drive->{iothread
};
1502 } elsif ($drive->{interface
} eq 'scsi') {
1504 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
1505 my $unit = $drive->{index} % $maxdev;
1506 my $devicetype = 'hd';
1508 if (drive_is_cdrom
($drive)) {
1511 if ($drive->{file
} =~ m
|^/|) {
1512 $path = $drive->{file
};
1513 if (my $info = path_is_scsi
($path)) {
1514 if ($info->{type
} == 0 && $drive->{scsiblock
}) {
1515 $devicetype = 'block';
1516 } elsif ($info->{type
} == 1) { # tape
1517 $devicetype = 'generic';
1521 $path = PVE
::Storage
::path
($storecfg, $drive->{file
});
1524 # for compatibility only, we prefer scsi-hd (#2408, #2355, #2380)
1525 my $version = extract_version
($machine_type, kvm_user_version
());
1526 if ($path =~ m/^iscsi\:\/\
// &&
1527 !min_version
($version, 4, 1)) {
1528 $devicetype = 'generic';
1532 if (!$conf->{scsihw
} || $conf->{scsihw
} =~ m/^lsi/ || $conf->{scsihw
} eq 'pvscsi') {
1533 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit";
1535 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0"
1536 .",lun=$drive->{index}";
1538 $device .= ",drive=drive-$drive_id,id=$drive_id";
1540 if ($drive->{ssd
} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1541 $device .= ",rotation_rate=1";
1543 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1545 } elsif ($drive->{interface
} eq 'ide' || $drive->{interface
} eq 'sata') {
1546 my $maxdev = ($drive->{interface
} eq 'sata') ?
$PVE::QemuServer
::Drive
::MAX_SATA_DISKS
: 2;
1547 my $controller = int($drive->{index} / $maxdev);
1548 my $unit = $drive->{index} % $maxdev;
1549 my $devicetype = ($drive->{media
} && $drive->{media
} eq 'cdrom') ?
"cd" : "hd";
1551 $device = "ide-$devicetype";
1552 if ($drive->{interface
} eq 'ide') {
1553 $device .= ",bus=ide.$controller,unit=$unit";
1555 $device .= ",bus=ahci$controller.$unit";
1557 $device .= ",drive=drive-$drive_id,id=$drive_id";
1559 if ($devicetype eq 'hd') {
1560 if (my $model = $drive->{model
}) {
1561 $model = URI
::Escape
::uri_unescape
($model);
1562 $device .= ",model=$model";
1564 if ($drive->{ssd
}) {
1565 $device .= ",rotation_rate=1";
1568 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn
};
1569 } elsif ($drive->{interface
} eq 'usb') {
1571 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1573 die "unsupported interface type";
1576 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex
};
1578 if (my $serial = $drive->{serial
}) {
1579 $serial = URI
::Escape
::uri_unescape
($serial);
1580 $device .= ",serial=$serial";
1587 sub get_initiator_name
{
1590 my $fh = IO
::File-
>new('/etc/iscsi/initiatorname.iscsi') || return;
1591 while (defined(my $line = <$fh>)) {
1592 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1601 my sub storage_allows_io_uring_default
{
1602 my ($scfg, $cache_direct) = @_;
1604 # io_uring with cache mode writeback or writethrough on krbd will hang...
1605 return if $scfg && $scfg->{type
} eq 'rbd' && $scfg->{krbd
} && !$cache_direct;
1607 # io_uring with cache mode writeback or writethrough on LVM will hang, without cache only
1608 # sometimes, just plain disable...
1609 return if $scfg && $scfg->{type
} eq 'lvm';
1611 # io_uring causes problems when used with CIFS since kernel 5.15
1612 # Some discussion: https://www.spinics.net/lists/linux-cifs/msg26734.html
1613 return if $scfg && $scfg->{type
} eq 'cifs';
1618 my sub drive_uses_cache_direct
{
1619 my ($drive, $scfg) = @_;
1621 my $cache_direct = 0;
1623 if (my $cache = $drive->{cache
}) {
1624 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1625 } elsif (!drive_is_cdrom
($drive) && !($scfg && $scfg->{type
} eq 'btrfs' && !$scfg->{nocow
})) {
1629 return $cache_direct;
1632 sub print_drive_commandline_full
{
1633 my ($storecfg, $vmid, $drive, $pbs_name, $io_uring) = @_;
1636 my $volid = $drive->{file
};
1637 my $format = $drive->{format
};
1638 my $drive_id = get_drive_id
($drive);
1640 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1641 my $scfg = $storeid ? PVE
::Storage
::storage_config
($storecfg, $storeid) : undef;
1643 if (drive_is_cdrom
($drive)) {
1644 $path = get_iso_path
($storecfg, $vmid, $volid);
1645 die "$drive_id: cannot back cdrom drive with PBS snapshot\n" if $pbs_name;
1648 $path = PVE
::Storage
::path
($storecfg, $volid);
1649 $format //= qemu_img_format
($scfg, $volname);
1656 my $is_rbd = $path =~ m/^rbd:/;
1659 my @qemu_drive_options = qw(heads secs cyls trans media cache rerror werror aio discard);
1660 foreach my $o (@qemu_drive_options) {
1661 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1664 # snapshot only accepts on|off
1665 if (defined($drive->{snapshot
})) {
1666 my $v = $drive->{snapshot
} ?
'on' : 'off';
1667 $opts .= ",snapshot=$v";
1670 if (defined($drive->{ro
})) { # ro maps to QEMUs `readonly`, which accepts `on` or `off` only
1671 $opts .= ",readonly=" . ($drive->{ro
} ?
'on' : 'off');
1674 foreach my $type (['', '-total'], [_rd
=> '-read'], [_wr
=> '-write']) {
1675 my ($dir, $qmpname) = @$type;
1676 if (my $v = $drive->{"mbps$dir"}) {
1677 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1679 if (my $v = $drive->{"mbps${dir}_max"}) {
1680 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1682 if (my $v = $drive->{"bps${dir}_max_length"}) {
1683 $opts .= ",throttling.bps$qmpname-max-length=$v";
1685 if (my $v = $drive->{"iops${dir}"}) {
1686 $opts .= ",throttling.iops$qmpname=$v";
1688 if (my $v = $drive->{"iops${dir}_max"}) {
1689 $opts .= ",throttling.iops$qmpname-max=$v";
1691 if (my $v = $drive->{"iops${dir}_max_length"}) {
1692 $opts .= ",throttling.iops$qmpname-max-length=$v";
1697 $format = "rbd" if $is_rbd;
1698 die "$drive_id: Proxmox Backup Server backed drive cannot auto-detect the format\n"
1700 $opts .= ",format=alloc-track,file.driver=$format";
1702 $opts .= ",format=$format";
1705 my $cache_direct = drive_uses_cache_direct
($drive, $scfg);
1707 $opts .= ",cache=none" if !$drive->{cache
} && $cache_direct;
1709 if (!$drive->{aio
}) {
1710 if ($io_uring && storage_allows_io_uring_default
($scfg, $cache_direct)) {
1711 # io_uring supports all cache modes
1712 $opts .= ",aio=io_uring";
1714 # aio native works only with O_DIRECT
1716 $opts .= ",aio=native";
1718 $opts .= ",aio=threads";
1723 if (!drive_is_cdrom
($drive)) {
1725 if (defined($drive->{detect_zeroes
}) && !$drive->{detect_zeroes
}) {
1726 $detectzeroes = 'off';
1727 } elsif ($drive->{discard
}) {
1728 $detectzeroes = $drive->{discard
} eq 'on' ?
'unmap' : 'on';
1730 # This used to be our default with discard not being specified:
1731 $detectzeroes = 'on';
1734 # note: 'detect-zeroes' works per blockdev and we want it to persist
1735 # after the alloc-track is removed, so put it on 'file' directly
1736 my $dz_param = $pbs_name ?
"file.detect-zeroes" : "detect-zeroes";
1737 $opts .= ",$dz_param=$detectzeroes" if $detectzeroes;
1741 $opts .= ",backing=$pbs_name";
1742 $opts .= ",auto-remove=on";
1745 # my $file_param = $pbs_name ? "file.file.filename" : "file";
1746 my $file_param = "file";
1748 # non-rbd drivers require the underlying file to be a seperate block
1749 # node, so add a second .file indirection
1750 $file_param .= ".file" if !$is_rbd;
1751 $file_param .= ".filename";
1753 my $pathinfo = $path ?
"$file_param=$path," : '';
1755 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1758 sub print_pbs_blockdev
{
1759 my ($pbs_conf, $pbs_name) = @_;
1760 my $blockdev = "driver=pbs,node-name=$pbs_name,read-only=on";
1761 $blockdev .= ",repository=$pbs_conf->{repository}";
1762 $blockdev .= ",namespace=$pbs_conf->{namespace}" if $pbs_conf->{namespace
};
1763 $blockdev .= ",snapshot=$pbs_conf->{snapshot}";
1764 $blockdev .= ",archive=$pbs_conf->{archive}";
1765 $blockdev .= ",keyfile=$pbs_conf->{keyfile}" if $pbs_conf->{keyfile
};
1769 sub print_netdevice_full
{
1770 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version) = @_;
1772 my $device = $net->{model
};
1773 if ($net->{model
} eq 'virtio') {
1774 $device = 'virtio-net-pci';
1777 my $pciaddr = print_pci_addr
("$netid", $bridges, $arch, $machine_type);
1778 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1779 if ($net->{queues
} && $net->{queues
} > 1 && $net->{model
} eq 'virtio'){
1780 # Consider we have N queues, the number of vectors needed is 2 * N + 2, i.e., one per in
1781 # and out of each queue plus one config interrupt and control vector queue
1782 my $vectors = $net->{queues
} * 2 + 2;
1783 $tmpstr .= ",vectors=$vectors,mq=on";
1784 if (min_version
($machine_version, 7, 1)) {
1785 $tmpstr .= ",packed=on";
1789 if (min_version
($machine_version, 7, 1) && $net->{model
} eq 'virtio'){
1790 $tmpstr .= ",rx_queue_size=1024,tx_queue_size=1024";
1793 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex
} ;
1795 if (my $mtu = $net->{mtu
}) {
1796 if ($net->{model
} eq 'virtio' && $net->{bridge
}) {
1797 my $bridge_mtu = PVE
::Network
::read_bridge_mtu
($net->{bridge
});
1800 } elsif ($mtu < 576) {
1801 die "netdev $netid: MTU '$mtu' is smaller than the IP minimum MTU '576'\n";
1802 } elsif ($mtu > $bridge_mtu) {
1803 die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
1805 $tmpstr .= ",host_mtu=$mtu";
1807 warn "WARN: netdev $netid: ignoring MTU '$mtu', not using VirtIO or no bridge configured.\n";
1811 if ($use_old_bios_files) {
1813 if ($device eq 'virtio-net-pci') {
1814 $romfile = 'pxe-virtio.rom';
1815 } elsif ($device eq 'e1000') {
1816 $romfile = 'pxe-e1000.rom';
1817 } elsif ($device eq 'e1000e') {
1818 $romfile = 'pxe-e1000e.rom';
1819 } elsif ($device eq 'ne2k') {
1820 $romfile = 'pxe-ne2k_pci.rom';
1821 } elsif ($device eq 'pcnet') {
1822 $romfile = 'pxe-pcnet.rom';
1823 } elsif ($device eq 'rtl8139') {
1824 $romfile = 'pxe-rtl8139.rom';
1826 $tmpstr .= ",romfile=$romfile" if $romfile;
1832 sub print_netdev_full
{
1833 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
1836 if ($netid =~ m/^net(\d+)$/) {
1840 die "got strange net id '$i'\n" if $i >= ${MAX_NETS
};
1842 my $ifname = "tap${vmid}i$i";
1844 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1845 die "interface name '$ifname' is too long (max 15 character)\n"
1846 if length($ifname) >= 16;
1848 my $vhostparam = '';
1849 if (is_native
($arch)) {
1850 $vhostparam = ',vhost=on' if kernel_has_vhost_net
() && $net->{model
} eq 'virtio';
1853 my $vmname = $conf->{name
} || "vm$vmid";
1856 my $script = $hotplug ?
"pve-bridge-hotplug" : "pve-bridge";
1858 if ($net->{bridge
}) {
1859 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script"
1860 .",downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1862 $netdev = "type=user,id=$netid,hostname=$vmname";
1865 $netdev .= ",queues=$net->{queues}" if ($net->{queues
} && $net->{model
} eq 'virtio');
1871 'cirrus' => 'cirrus-vga',
1873 'vmware' => 'vmware-svga',
1874 'virtio' => 'virtio-vga',
1875 'virtio-gl' => 'virtio-vga-gl',
1878 sub print_vga_device
{
1879 my ($conf, $vga, $arch, $machine_version, $machine, $id, $qxlnum, $bridges) = @_;
1881 my $type = $vga_map->{$vga->{type
}};
1882 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
1883 $type = 'virtio-gpu';
1885 my $vgamem_mb = $vga->{memory
};
1887 my $max_outputs = '';
1889 $type = $id ?
'qxl' : 'qxl-vga';
1891 if (!$conf->{ostype
} || $conf->{ostype
} =~ m/^(?:l\d\d)|(?:other)$/) {
1892 # set max outputs so linux can have up to 4 qxl displays with one device
1893 if (min_version
($machine_version, 4, 1)) {
1894 $max_outputs = ",max_outputs=4";
1899 die "no devicetype for $vga->{type}\n" if !$type;
1903 if ($vga->{type
} =~ /^virtio/) {
1904 my $bytes = PVE
::Tools
::convert_size
($vgamem_mb, "mb" => "b");
1905 $memory = ",max_hostmem=$bytes";
1907 # from https://www.spice-space.org/multiple-monitors.html
1908 $memory = ",vgamem_mb=$vga->{memory}";
1909 my $ram = $vgamem_mb * 4;
1910 my $vram = $vgamem_mb * 2;
1911 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
1913 $memory = ",vgamem_mb=$vga->{memory}";
1915 } elsif ($qxlnum && $id) {
1916 $memory = ",ram_size=67108864,vram_size=33554432";
1920 if ($type eq 'VGA' && windows_version
($conf->{ostype
})) {
1921 $edidoff=",edid=off" if (!defined($conf->{bios
}) || $conf->{bios
} ne 'ovmf');
1924 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1925 my $vgaid = "vga" . ($id // '');
1927 if ($q35 && $vgaid eq 'vga') {
1928 # the first display uses pcie.0 bus on q35 machines
1929 $pciaddr = print_pcie_addr
($vgaid, $bridges, $arch, $machine);
1931 $pciaddr = print_pci_addr
($vgaid, $bridges, $arch, $machine);
1934 if ($vga->{type
} eq 'virtio-gl') {
1935 my $base = '/usr/lib/x86_64-linux-gnu/lib';
1936 die "missing libraries for '$vga->{type}' detected! Please install 'libgl1' and 'libegl1'\n"
1937 if !-e
"${base}EGL.so.1" || !-e
"${base}GL.so.1";
1939 die "no DRM render node detected (/dev/dri/renderD*), no GPU? - needed for '$vga->{type}' display\n"
1940 if !PVE
::Tools
::dir_glob_regex
('/dev/dri/', "renderD.*");
1943 return "$type,id=${vgaid}${memory}${max_outputs}${pciaddr}${edidoff}";
1946 sub parse_number_sets
{
1949 foreach my $part (split(/;/, $set)) {
1950 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1951 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1952 push @$res, [ $1, $2 ];
1954 die "invalid range: $part\n";
1963 my $res = parse_property_string
($numa_fmt, $data);
1964 $res->{cpus
} = parse_number_sets
($res->{cpus
}) if defined($res->{cpus
});
1965 $res->{hostnodes
} = parse_number_sets
($res->{hostnodes
}) if defined($res->{hostnodes
});
1969 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1971 my ($data, $disable_mac_autogen) = @_;
1973 my $res = eval { parse_property_string
($net_fmt, $data) };
1978 if (!defined($res->{macaddr
}) && !$disable_mac_autogen) {
1979 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
1980 $res->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
});
1985 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
1986 sub parse_ipconfig
{
1989 my $res = eval { parse_property_string
($ipconfig_fmt, $data) };
1995 if ($res->{gw
} && !$res->{ip
}) {
1996 warn 'gateway specified without specifying an IP address';
1999 if ($res->{gw6
} && !$res->{ip6
}) {
2000 warn 'IPv6 gateway specified without specifying an IPv6 address';
2003 if ($res->{gw
} && $res->{ip
} eq 'dhcp') {
2004 warn 'gateway specified together with DHCP';
2007 if ($res->{gw6
} && $res->{ip6
} !~ /^$IPV6RE/) {
2009 warn "IPv6 gateway specified together with $res->{ip6} address";
2013 if (!$res->{ip
} && !$res->{ip6
}) {
2014 return { ip
=> 'dhcp', ip6
=> 'dhcp' };
2023 return PVE
::JSONSchema
::print_property_string
($net, $net_fmt);
2026 sub add_random_macs
{
2027 my ($settings) = @_;
2029 foreach my $opt (keys %$settings) {
2030 next if $opt !~ m/^net(\d+)$/;
2031 my $net = parse_net
($settings->{$opt});
2033 $settings->{$opt} = print_net
($net);
2037 sub vm_is_volid_owner
{
2038 my ($storecfg, $vmid, $volid) = @_;
2040 if ($volid !~ m
|^/|) {
2042 eval { ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid); };
2043 if ($owner && ($owner == $vmid)) {
2051 sub vmconfig_register_unused_drive
{
2052 my ($storecfg, $vmid, $conf, $drive) = @_;
2054 if (drive_is_cloudinit
($drive)) {
2055 eval { PVE
::Storage
::vdisk_free
($storecfg, $drive->{file
}) };
2057 delete $conf->{cloudinit
};
2058 } elsif (!drive_is_cdrom
($drive)) {
2059 my $volid = $drive->{file
};
2060 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
2061 PVE
::QemuConfig-
>add_unused_volume($conf, $volid, $vmid);
2066 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2070 pattern
=> '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2071 format_description
=> 'UUID',
2072 description
=> "Set SMBIOS1 UUID.",
2077 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2078 format_description
=> 'Base64 encoded string',
2079 description
=> "Set SMBIOS1 version.",
2084 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2085 format_description
=> 'Base64 encoded string',
2086 description
=> "Set SMBIOS1 serial number.",
2091 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2092 format_description
=> 'Base64 encoded string',
2093 description
=> "Set SMBIOS1 manufacturer.",
2098 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2099 format_description
=> 'Base64 encoded string',
2100 description
=> "Set SMBIOS1 product ID.",
2105 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2106 format_description
=> 'Base64 encoded string',
2107 description
=> "Set SMBIOS1 SKU string.",
2112 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2113 format_description
=> 'Base64 encoded string',
2114 description
=> "Set SMBIOS1 family string.",
2119 description
=> 'Flag to indicate that the SMBIOS values are base64 encoded',
2127 my $res = eval { parse_property_string
($smbios1_fmt, $data) };
2134 return PVE
::JSONSchema
::print_property_string
($smbios1, $smbios1_fmt);
2137 PVE
::JSONSchema
::register_format
('pve-qm-smbios1', $smbios1_fmt);
2139 sub parse_watchdog
{
2144 my $res = eval { parse_property_string
($watchdog_fmt, $value) };
2149 sub parse_guest_agent
{
2152 return {} if !defined($conf->{agent
});
2154 my $res = eval { parse_property_string
($agent_fmt, $conf->{agent
}) };
2157 # if the agent is disabled ignore the other potentially set properties
2158 return {} if !$res->{enabled
};
2163 my ($conf, $key) = @_;
2164 return undef if !defined($conf->{agent
});
2166 my $agent = parse_guest_agent
($conf);
2167 return $agent->{$key};
2173 return {} if !$value;
2174 my $res = eval { parse_property_string
($vga_fmt, $value) };
2184 my $res = eval { parse_property_string
($rng_fmt, $value) };
2189 sub parse_meta_info
{
2194 my $res = eval { parse_property_string
($meta_info_fmt, $value) };
2199 sub new_meta_info_string
{
2200 my () = @_; # for now do not allow to override any value
2202 return PVE
::JSONSchema
::print_property_string
(
2204 'creation-qemu' => kvm_user_version
(),
2205 ctime
=> "". int(time()),
2211 sub qemu_created_version_fixups
{
2212 my ($conf, $forcemachine, $kvmver) = @_;
2214 my $meta = parse_meta_info
($conf->{meta
}) // {};
2215 my $forced_vers = PVE
::QemuServer
::Machine
::extract_version
($forcemachine);
2217 # check if we need to apply some handling for VMs that always use the latest machine version but
2218 # had a machine version transition happen that affected HW such that, e.g., an OS config change
2219 # would be required (we do not want to pin machine version for non-windows OS type)
2221 (!defined($conf->{machine
}) || $conf->{machine
} =~ m/^(?:pc|q35|virt)$/) # non-versioned machine
2222 && (!defined($meta->{'creation-qemu'}) || !min_version
($meta->{'creation-qemu'}, 6, 1)) # created before 6.1
2223 && (!$forced_vers || min_version
($forced_vers, 6, 1)) # handle snapshot-rollback/migrations
2224 && min_version
($kvmver, 6, 1) # only need to apply the change since 6.1
2226 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
2227 if ($q35 && $conf->{ostype
} && $conf->{ostype
} eq 'l26') {
2228 # this changed to default-on in Q 6.1 for q35 machines, it will mess with PCI slot view
2229 # and thus with the predictable interface naming of systemd
2230 return ['-global', 'ICH9-LPC.acpi-pci-hotplug-with-bridge-support=off'];
2236 PVE
::JSONSchema
::register_format
('pve-qm-usb-device', \
&verify_usb_device
);
2237 sub verify_usb_device
{
2238 my ($value, $noerr) = @_;
2240 return $value if parse_usb_device
($value);
2244 die "unable to parse usb device\n";
2247 # add JSON properties for create and set function
2248 sub json_config_properties
{
2249 my ($prop, $with_disk_alloc) = @_;
2251 my $skip_json_config_opts = {
2255 runningmachine
=> 1,
2260 foreach my $opt (keys %$confdesc) {
2261 next if $skip_json_config_opts->{$opt};
2263 if ($with_disk_alloc && is_valid_drivename
($opt)) {
2264 $prop->{$opt} = $PVE::QemuServer
::Drive
::drivedesc_hash_with_alloc-
>{$opt};
2266 $prop->{$opt} = $confdesc->{$opt};
2273 # Properties that we can read from an OVF file
2274 sub json_ovf_properties
{
2277 for my $device (PVE
::QemuServer
::Drive
::valid_drive_names
()) {
2278 $prop->{$device} = {
2280 format
=> 'pve-volume-id-or-absolute-path',
2281 description
=> "Disk image that gets imported to $device",
2288 description
=> "The number of CPU cores.",
2293 description
=> "Amount of RAM for the VM in MB.",
2298 description
=> "Name of the VM.",
2305 # return copy of $confdesc_cloudinit to generate documentation
2306 sub cloudinit_config_properties
{
2308 return dclone
($confdesc_cloudinit);
2311 sub cloudinit_pending_properties
{
2313 map { $_ => 1 } keys $confdesc_cloudinit->%*,
2316 $p->{"net$_"} = 1 for 0..($MAX_NETS-1);
2321 my ($key, $value) = @_;
2323 die "unknown setting '$key'\n" if !$confdesc->{$key};
2325 my $type = $confdesc->{$key}->{type
};
2327 if (!defined($value)) {
2328 die "got undefined value\n";
2331 if ($value =~ m/[\n\r]/) {
2332 die "property contains a line feed\n";
2335 if ($type eq 'boolean') {
2336 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2337 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2338 die "type check ('boolean') failed - got '$value'\n";
2339 } elsif ($type eq 'integer') {
2340 return int($1) if $value =~ m/^(\d+)$/;
2341 die "type check ('integer') failed - got '$value'\n";
2342 } elsif ($type eq 'number') {
2343 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2344 die "type check ('number') failed - got '$value'\n";
2345 } elsif ($type eq 'string') {
2346 if (my $fmt = $confdesc->{$key}->{format
}) {
2347 PVE
::JSONSchema
::check_format
($fmt, $value);
2350 $value =~ s/^\"(.*)\"$/$1/;
2353 die "internal error"
2358 my ($storecfg, $vmid, $skiplock, $replacement_conf, $purge_unreferenced) = @_;
2360 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2362 if (!$skiplock && !PVE
::QemuConfig-
>has_lock($conf, 'suspended')) {
2363 PVE
::QemuConfig-
>check_lock($conf);
2366 if ($conf->{template
}) {
2367 # check if any base image is still used by a linked clone
2368 PVE
::QemuConfig-
>foreach_volume_full($conf, { include_unused
=> 1 }, sub {
2369 my ($ds, $drive) = @_;
2370 return if drive_is_cdrom
($drive);
2372 my $volid = $drive->{file
};
2373 return if !$volid || $volid =~ m
|^/|;
2375 die "base volume '$volid' is still in use by linked cloned\n"
2376 if PVE
::Storage
::volume_is_base_and_used
($storecfg, $volid);
2382 my $remove_owned_drive = sub {
2383 my ($ds, $drive) = @_;
2384 return if drive_is_cdrom
($drive, 1);
2386 my $volid = $drive->{file
};
2387 return if !$volid || $volid =~ m
|^/|;
2388 return if $volids->{$volid};
2390 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
2391 return if !$path || !$owner || ($owner != $vmid);
2393 $volids->{$volid} = 1;
2394 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2395 warn "Could not remove disk '$volid', check manually: $@" if $@;
2398 # only remove disks owned by this VM (referenced in the config)
2399 my $include_opts = {
2400 include_unused
=> 1,
2401 extra_keys
=> ['vmstate'],
2403 PVE
::QemuConfig-
>foreach_volume_full($conf, $include_opts, $remove_owned_drive);
2405 for my $snap (values %{$conf->{snapshots
}}) {
2406 next if !defined($snap->{vmstate
});
2407 my $drive = PVE
::QemuConfig-
>parse_volume('vmstate', $snap->{vmstate
}, 1);
2408 next if !defined($drive);
2409 $remove_owned_drive->('vmstate', $drive);
2412 PVE
::QemuConfig-
>foreach_volume_full($conf->{pending
}, $include_opts, $remove_owned_drive);
2414 if ($purge_unreferenced) { # also remove unreferenced disk
2415 my $vmdisks = PVE
::Storage
::vdisk_list
($storecfg, undef, $vmid, undef, 'images');
2416 PVE
::Storage
::foreach_volid
($vmdisks, sub {
2417 my ($volid, $sid, $volname, $d) = @_;
2418 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2423 if (defined $replacement_conf) {
2424 PVE
::QemuConfig-
>write_config($vmid, $replacement_conf);
2426 PVE
::QemuConfig-
>destroy_config($vmid);
2430 sub parse_vm_config
{
2431 my ($filename, $raw, $strict) = @_;
2433 return if !defined($raw);
2436 digest
=> Digest
::SHA
::sha1_hex
($raw),
2442 my $handle_error = sub {
2452 $filename =~ m
|/qemu-server/(\d
+)\
.conf
$|
2453 || die "got strange filename '$filename'";
2459 my $finish_description = sub {
2460 if (defined($descr)) {
2462 $conf->{description
} = $descr;
2468 my @lines = split(/\n/, $raw);
2469 foreach my $line (@lines) {
2470 next if $line =~ m/^\s*$/;
2472 if ($line =~ m/^\[PENDING\]\s*$/i) {
2473 $section = 'pending';
2474 $finish_description->();
2475 $conf = $res->{$section} = {};
2477 } elsif ($line =~ m/^\[special:cloudinit\]\s*$/i) {
2478 $section = 'cloudinit';
2479 $finish_description->();
2480 $conf = $res->{$section} = {};
2483 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2485 $finish_description->();
2486 $conf = $res->{snapshots
}->{$section} = {};
2490 if ($line =~ m/^\#(.*)$/) {
2491 $descr = '' if !defined($descr);
2492 $descr .= PVE
::Tools
::decode_text
($1) . "\n";
2496 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2497 $descr = '' if !defined($descr);
2498 $descr .= PVE
::Tools
::decode_text
($2);
2499 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2500 $conf->{snapstate
} = $1;
2501 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2504 $conf->{$key} = $value;
2505 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2507 if ($section eq 'pending') {
2508 $conf->{delete} = $value; # we parse this later
2510 $handle_error->("vm $vmid - property 'delete' is only allowed in [PENDING]\n");
2512 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2515 if ($section eq 'cloudinit') {
2516 # ignore validation only used for informative purpose
2517 $conf->{$key} = $value;
2520 eval { $value = check_type
($key, $value); };
2522 $handle_error->("vm $vmid - unable to parse value of '$key' - $@");
2524 $key = 'ide2' if $key eq 'cdrom';
2525 my $fmt = $confdesc->{$key}->{format
};
2526 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2527 my $v = parse_drive
($key, $value);
2528 if (my $volid = filename_to_volume_id
($vmid, $v->{file
}, $v->{media
})) {
2529 $v->{file
} = $volid;
2530 $value = print_drive
($v);
2532 $handle_error->("vm $vmid - unable to parse value of '$key'\n");
2537 $conf->{$key} = $value;
2540 $handle_error->("vm $vmid - unable to parse config: $line\n");
2544 $finish_description->();
2545 delete $res->{snapstate
}; # just to be sure
2550 sub write_vm_config
{
2551 my ($filename, $conf) = @_;
2553 delete $conf->{snapstate
}; # just to be sure
2555 if ($conf->{cdrom
}) {
2556 die "option ide2 conflicts with cdrom\n" if $conf->{ide2
};
2557 $conf->{ide2
} = $conf->{cdrom
};
2558 delete $conf->{cdrom
};
2561 # we do not use 'smp' any longer
2562 if ($conf->{sockets
}) {
2563 delete $conf->{smp
};
2564 } elsif ($conf->{smp
}) {
2565 $conf->{sockets
} = $conf->{smp
};
2566 delete $conf->{cores
};
2567 delete $conf->{smp
};
2570 my $used_volids = {};
2572 my $cleanup_config = sub {
2573 my ($cref, $pending, $snapname) = @_;
2575 foreach my $key (keys %$cref) {
2576 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2577 $key eq 'snapstate' || $key eq 'pending' || $key eq 'cloudinit';
2578 my $value = $cref->{$key};
2579 if ($key eq 'delete') {
2580 die "propertry 'delete' is only allowed in [PENDING]\n"
2582 # fixme: check syntax?
2585 eval { $value = check_type
($key, $value); };
2586 die "unable to parse value of '$key' - $@" if $@;
2588 $cref->{$key} = $value;
2590 if (!$snapname && is_valid_drivename
($key)) {
2591 my $drive = parse_drive
($key, $value);
2592 $used_volids->{$drive->{file
}} = 1 if $drive && $drive->{file
};
2597 &$cleanup_config($conf);
2599 &$cleanup_config($conf->{pending
}, 1);
2601 foreach my $snapname (keys %{$conf->{snapshots
}}) {
2602 die "internal error: snapshot name '$snapname' is forbidden" if lc($snapname) eq 'pending';
2603 &$cleanup_config($conf->{snapshots
}->{$snapname}, undef, $snapname);
2606 # remove 'unusedX' settings if we re-add a volume
2607 foreach my $key (keys %$conf) {
2608 my $value = $conf->{$key};
2609 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2610 delete $conf->{$key};
2614 my $generate_raw_config = sub {
2615 my ($conf, $pending) = @_;
2619 # add description as comment to top of file
2620 if (defined(my $descr = $conf->{description
})) {
2622 foreach my $cl (split(/\n/, $descr)) {
2623 $raw .= '#' . PVE
::Tools
::encode_text
($cl) . "\n";
2626 $raw .= "#\n" if $pending;
2630 foreach my $key (sort keys %$conf) {
2631 next if $key =~ /^(digest|description|pending|cloudinit|snapshots)$/;
2632 $raw .= "$key: $conf->{$key}\n";
2637 my $raw = &$generate_raw_config($conf);
2639 if (scalar(keys %{$conf->{pending
}})){
2640 $raw .= "\n[PENDING]\n";
2641 $raw .= &$generate_raw_config($conf->{pending
}, 1);
2644 if (scalar(keys %{$conf->{cloudinit
}}) && PVE
::QemuConfig-
>has_cloudinit($conf)){
2645 $raw .= "\n[special:cloudinit]\n";
2646 $raw .= &$generate_raw_config($conf->{cloudinit
});
2649 foreach my $snapname (sort keys %{$conf->{snapshots
}}) {
2650 $raw .= "\n[$snapname]\n";
2651 $raw .= &$generate_raw_config($conf->{snapshots
}->{$snapname});
2661 # we use static defaults from our JSON schema configuration
2662 foreach my $key (keys %$confdesc) {
2663 if (defined(my $default = $confdesc->{$key}->{default})) {
2664 $res->{$key} = $default;
2672 my $vmlist = PVE
::Cluster
::get_vmlist
();
2674 return $res if !$vmlist || !$vmlist->{ids
};
2675 my $ids = $vmlist->{ids
};
2676 my $nodename = nodename
();
2678 foreach my $vmid (keys %$ids) {
2679 my $d = $ids->{$vmid};
2680 next if !$d->{node
} || $d->{node
} ne $nodename;
2681 next if !$d->{type
} || $d->{type
} ne 'qemu';
2682 $res->{$vmid}->{exists} = 1;
2687 # test if VM uses local resources (to prevent migration)
2688 sub check_local_resources
{
2689 my ($conf, $noerr) = @_;
2693 push @loc_res, "hostusb" if $conf->{hostusb
}; # old syntax
2694 push @loc_res, "hostpci" if $conf->{hostpci
}; # old syntax
2696 push @loc_res, "ivshmem" if $conf->{ivshmem
};
2698 foreach my $k (keys %$conf) {
2699 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2700 # sockets are safe: they will recreated be on the target side post-migrate
2701 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2702 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2705 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2710 # check if used storages are available on all nodes (use by migrate)
2711 sub check_storage_availability
{
2712 my ($storecfg, $conf, $node) = @_;
2714 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2715 my ($ds, $drive) = @_;
2717 my $volid = $drive->{file
};
2720 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2723 # check if storage is available on both nodes
2724 my $scfg = PVE
::Storage
::storage_check_enabled
($storecfg, $sid);
2725 PVE
::Storage
::storage_check_enabled
($storecfg, $sid, $node);
2727 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $volid);
2729 die "$volid: content type '$vtype' is not available on storage '$sid'\n"
2730 if !$scfg->{content
}->{$vtype};
2734 # list nodes where all VM images are available (used by has_feature API)
2736 my ($conf, $storecfg) = @_;
2738 my $nodelist = PVE
::Cluster
::get_nodelist
();
2739 my $nodehash = { map { $_ => 1 } @$nodelist };
2740 my $nodename = nodename
();
2742 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2743 my ($ds, $drive) = @_;
2745 my $volid = $drive->{file
};
2748 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2750 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2751 if ($scfg->{disable
}) {
2753 } elsif (my $avail = $scfg->{nodes
}) {
2754 foreach my $node (keys %$nodehash) {
2755 delete $nodehash->{$node} if !$avail->{$node};
2757 } elsif (!$scfg->{shared
}) {
2758 foreach my $node (keys %$nodehash) {
2759 delete $nodehash->{$node} if $node ne $nodename
2768 sub check_local_storage_availability
{
2769 my ($conf, $storecfg) = @_;
2771 my $nodelist = PVE
::Cluster
::get_nodelist
();
2772 my $nodehash = { map { $_ => {} } @$nodelist };
2774 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2775 my ($ds, $drive) = @_;
2777 my $volid = $drive->{file
};
2780 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2782 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2784 if ($scfg->{disable
}) {
2785 foreach my $node (keys %$nodehash) {
2786 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2788 } elsif (my $avail = $scfg->{nodes
}) {
2789 foreach my $node (keys %$nodehash) {
2790 if (!$avail->{$node}) {
2791 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2798 foreach my $node (values %$nodehash) {
2799 if (my $unavail = $node->{unavailable_storages
}) {
2800 $node->{unavailable_storages
} = [ sort keys %$unavail ];
2807 # Compat only, use assert_config_exists_on_node and vm_running_locally where possible
2809 my ($vmid, $nocheck, $node) = @_;
2811 # $nocheck is set when called during a migration, in which case the config
2812 # file might still or already reside on the *other* node
2813 # - because rename has already happened, and current node is source
2814 # - because rename hasn't happened yet, and current node is target
2815 # - because rename has happened, current node is target, but hasn't yet
2817 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid, $node) if !$nocheck;
2818 return PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
2823 my $vzlist = config_list
();
2825 my $fd = IO
::Dir-
>new($PVE::QemuServer
::Helpers
::var_run_tmpdir
) || return $vzlist;
2827 while (defined(my $de = $fd->read)) {
2828 next if $de !~ m/^(\d+)\.pid$/;
2830 next if !defined($vzlist->{$vmid});
2831 if (my $pid = check_running
($vmid)) {
2832 $vzlist->{$vmid}->{pid
} = $pid;
2839 our $vmstatus_return_properties = {
2840 vmid
=> get_standard_option
('pve-vmid'),
2842 description
=> "QEMU process status.",
2844 enum
=> ['stopped', 'running'],
2847 description
=> "Maximum memory in bytes.",
2850 renderer
=> 'bytes',
2853 description
=> "Root disk size in bytes.",
2856 renderer
=> 'bytes',
2859 description
=> "VM name.",
2864 description
=> "QEMU QMP agent status.",
2869 description
=> "PID of running qemu process.",
2874 description
=> "Uptime.",
2877 renderer
=> 'duration',
2880 description
=> "Maximum usable CPUs.",
2885 description
=> "The current config lock, if any.",
2890 description
=> "The current configured tags, if any",
2894 'running-machine' => {
2895 description
=> "The currently running machine type (if running).",
2900 description
=> "The currently running QEMU version (if running).",
2906 my $last_proc_pid_stat;
2908 # get VM status information
2909 # This must be fast and should not block ($full == false)
2910 # We only query KVM using QMP if $full == true (this can be slow)
2912 my ($opt_vmid, $full) = @_;
2916 my $storecfg = PVE
::Storage
::config
();
2918 my $list = vzlist
();
2919 my $defaults = load_defaults
();
2921 my ($uptime) = PVE
::ProcFSTools
::read_proc_uptime
(1);
2923 my $cpucount = $cpuinfo->{cpus
} || 1;
2925 foreach my $vmid (keys %$list) {
2926 next if $opt_vmid && ($vmid ne $opt_vmid);
2928 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2930 my $d = { vmid
=> int($vmid) };
2931 $d->{pid
} = int($list->{$vmid}->{pid
}) if $list->{$vmid}->{pid
};
2933 # fixme: better status?
2934 $d->{status
} = $list->{$vmid}->{pid
} ?
'running' : 'stopped';
2936 my $size = PVE
::QemuServer
::Drive
::bootdisk_size
($storecfg, $conf);
2937 if (defined($size)) {
2938 $d->{disk
} = 0; # no info available
2939 $d->{maxdisk
} = $size;
2945 $d->{cpus
} = ($conf->{sockets
} || $defaults->{sockets
})
2946 * ($conf->{cores
} || $defaults->{cores
});
2947 $d->{cpus
} = $cpucount if $d->{cpus
} > $cpucount;
2948 $d->{cpus
} = $conf->{vcpus
} if $conf->{vcpus
};
2950 $d->{name
} = $conf->{name
} || "VM $vmid";
2951 $d->{maxmem
} = $conf->{memory
} ?
$conf->{memory
}*(1024*1024)
2952 : $defaults->{memory
}*(1024*1024);
2954 if ($conf->{balloon
}) {
2955 $d->{balloon_min
} = $conf->{balloon
}*(1024*1024);
2956 $d->{shares
} = defined($conf->{shares
}) ?
$conf->{shares
}
2957 : $defaults->{shares
};
2968 $d->{diskwrite
} = 0;
2970 $d->{template
} = 1 if PVE
::QemuConfig-
>is_template($conf);
2972 $d->{serial
} = 1 if conf_has_serial
($conf);
2973 $d->{lock} = $conf->{lock} if $conf->{lock};
2974 $d->{tags
} = $conf->{tags
} if defined($conf->{tags
});
2979 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
2980 foreach my $dev (keys %$netdev) {
2981 next if $dev !~ m/^tap([1-9]\d*)i/;
2983 my $d = $res->{$vmid};
2986 $d->{netout
} += $netdev->{$dev}->{receive
};
2987 $d->{netin
} += $netdev->{$dev}->{transmit
};
2990 $d->{nics
}->{$dev}->{netout
} = int($netdev->{$dev}->{receive
});
2991 $d->{nics
}->{$dev}->{netin
} = int($netdev->{$dev}->{transmit
});
2996 my $ctime = gettimeofday
;
2998 foreach my $vmid (keys %$list) {
3000 my $d = $res->{$vmid};
3001 my $pid = $d->{pid
};
3004 my $pstat = PVE
::ProcFSTools
::read_proc_pid_stat
($pid);
3005 next if !$pstat; # not running
3007 my $used = $pstat->{utime} + $pstat->{stime
};
3009 $d->{uptime
} = int(($uptime - $pstat->{starttime
})/$cpuinfo->{user_hz
});
3011 if ($pstat->{vsize
}) {
3012 $d->{mem
} = int(($pstat->{rss
}/$pstat->{vsize
})*$d->{maxmem
});
3015 my $old = $last_proc_pid_stat->{$pid};
3017 $last_proc_pid_stat->{$pid} = {
3025 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
3027 if ($dtime > 1000) {
3028 my $dutime = $used - $old->{used
};
3030 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
3031 $last_proc_pid_stat->{$pid} = {
3037 $d->{cpu
} = $old->{cpu
};
3041 return $res if !$full;
3043 my $qmpclient = PVE
::QMPClient-
>new();
3045 my $ballooncb = sub {
3046 my ($vmid, $resp) = @_;
3048 my $info = $resp->{'return'};
3049 return if !$info->{max_mem
};
3051 my $d = $res->{$vmid};
3053 # use memory assigned to VM
3054 $d->{maxmem
} = $info->{max_mem
};
3055 $d->{balloon
} = $info->{actual
};
3057 if (defined($info->{total_mem
}) && defined($info->{free_mem
})) {
3058 $d->{mem
} = $info->{total_mem
} - $info->{free_mem
};
3059 $d->{freemem
} = $info->{free_mem
};
3062 $d->{ballooninfo
} = $info;
3065 my $blockstatscb = sub {
3066 my ($vmid, $resp) = @_;
3067 my $data = $resp->{'return'} || [];
3068 my $totalrdbytes = 0;
3069 my $totalwrbytes = 0;
3071 for my $blockstat (@$data) {
3072 $totalrdbytes = $totalrdbytes + $blockstat->{stats
}->{rd_bytes
};
3073 $totalwrbytes = $totalwrbytes + $blockstat->{stats
}->{wr_bytes
};
3075 $blockstat->{device
} =~ s/drive-//;
3076 $res->{$vmid}->{blockstat
}->{$blockstat->{device
}} = $blockstat->{stats
};
3078 $res->{$vmid}->{diskread
} = $totalrdbytes;
3079 $res->{$vmid}->{diskwrite
} = $totalwrbytes;
3082 my $machinecb = sub {
3083 my ($vmid, $resp) = @_;
3084 my $data = $resp->{'return'} || [];
3086 $res->{$vmid}->{'running-machine'} =
3087 PVE
::QemuServer
::Machine
::current_from_query_machines
($data);
3090 my $versioncb = sub {
3091 my ($vmid, $resp) = @_;
3092 my $data = $resp->{'return'} // {};
3093 my $version = 'unknown';
3095 if (my $v = $data->{qemu
}) {
3096 $version = $v->{major
} . "." . $v->{minor
} . "." . $v->{micro
};
3099 $res->{$vmid}->{'running-qemu'} = $version;
3102 my $statuscb = sub {
3103 my ($vmid, $resp) = @_;
3105 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3106 $qmpclient->queue_cmd($vmid, $machinecb, 'query-machines');
3107 $qmpclient->queue_cmd($vmid, $versioncb, 'query-version');
3108 # this fails if ballon driver is not loaded, so this must be
3109 # the last commnand (following command are aborted if this fails).
3110 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3112 my $status = 'unknown';
3113 if (!defined($status = $resp->{'return'}->{status
})) {
3114 warn "unable to get VM status\n";
3118 $res->{$vmid}->{qmpstatus
} = $resp->{'return'}->{status
};
3121 foreach my $vmid (keys %$list) {
3122 next if $opt_vmid && ($vmid ne $opt_vmid);
3123 next if !$res->{$vmid}->{pid
}; # not running
3124 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3127 $qmpclient->queue_execute(undef, 2);
3129 foreach my $vmid (keys %$list) {
3130 next if $opt_vmid && ($vmid ne $opt_vmid);
3131 next if !$res->{$vmid}->{pid
}; #not running
3133 # we can't use the $qmpclient since it might have already aborted on
3134 # 'query-balloon', but this might also fail for older versions...
3135 my $qemu_support = eval { mon_cmd
($vmid, "query-proxmox-support") };
3136 $res->{$vmid}->{'proxmox-support'} = $qemu_support // {};
3139 foreach my $vmid (keys %$list) {
3140 next if $opt_vmid && ($vmid ne $opt_vmid);
3141 $res->{$vmid}->{qmpstatus
} = $res->{$vmid}->{status
} if !$res->{$vmid}->{qmpstatus
};
3147 sub conf_has_serial
{
3150 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3151 if ($conf->{"serial$i"}) {
3159 sub conf_has_audio
{
3160 my ($conf, $id) = @_;
3163 my $audio = $conf->{"audio$id"};
3164 return if !defined($audio);
3166 my $audioproperties = parse_property_string
($audio_fmt, $audio);
3167 my $audiodriver = $audioproperties->{driver
} // 'spice';
3170 dev
=> $audioproperties->{device
},
3171 dev_id
=> "audiodev$id",
3172 backend
=> $audiodriver,
3173 backend_id
=> "$audiodriver-backend${id}",
3178 my ($audio, $audiopciaddr, $machine_version) = @_;
3182 my $id = $audio->{dev_id
};
3184 if (min_version
($machine_version, 4, 2)) {
3185 $audiodev = ",audiodev=$audio->{backend_id}";
3188 if ($audio->{dev
} eq 'AC97') {
3189 push @$devs, '-device', "AC97,id=${id}${audiopciaddr}$audiodev";
3190 } elsif ($audio->{dev
} =~ /intel\-hda$/) {
3191 push @$devs, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3192 push @$devs, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0$audiodev";
3193 push @$devs, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1$audiodev";
3195 die "unkown audio device '$audio->{dev}', implement me!";
3198 push @$devs, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3206 socket => "/var/run/qemu-server/$vmid.swtpm",
3207 pid
=> "/var/run/qemu-server/$vmid.swtpm.pid",
3211 sub add_tpm_device
{
3212 my ($vmid, $devices, $conf) = @_;
3214 return if !$conf->{tpmstate0
};
3216 my $paths = get_tpm_paths
($vmid);
3218 push @$devices, "-chardev", "socket,id=tpmchar,path=$paths->{socket}";
3219 push @$devices, "-tpmdev", "emulator,id=tpmdev,chardev=tpmchar";
3220 push @$devices, "-device", "tpm-tis,tpmdev=tpmdev";
3224 my ($storecfg, $vmid, $tpmdrive, $migration) = @_;
3226 return if !$tpmdrive;
3229 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
3230 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
3232 $state = PVE
::Storage
::map_volume
($storecfg, $tpm->{file
});
3234 $state = $tpm->{file
};
3237 my $paths = get_tpm_paths
($vmid);
3239 # during migration, we will get state from remote
3242 # run swtpm_setup to create a new TPM state if it doesn't exist yet
3249 "--create-platform-cert",
3252 "/etc/swtpm_setup.conf", # do not use XDG configs
3254 "0", # force creation as root, error if not possible
3255 "--not-overwrite", # ignore existing state, do not modify
3258 push @$setup_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3259 # TPM 2.0 supports ECC crypto, use if possible
3260 push @$setup_cmd, "--ecc" if $tpm->{version
} eq 'v2.0';
3262 run_command
($setup_cmd, outfunc
=> sub {
3263 print "swtpm_setup: $1\n";
3267 my $emulator_cmd = [
3271 "backend-uri=file://$state,mode=0600",
3273 "type=unixio,path=$paths->{socket},mode=0600",
3275 "file=$paths->{pid}",
3276 "--terminate", # terminate on QEMU disconnect
3279 push @$emulator_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3280 run_command
($emulator_cmd, outfunc
=> sub { print $1; });
3282 my $tries = 100; # swtpm may take a bit to start before daemonizing, wait up to 5s for pid
3283 while (! -e
$paths->{pid
}) {
3284 die "failed to start swtpm: pid file '$paths->{pid}' wasn't created.\n" if --$tries == 0;
3288 # return untainted PID of swtpm daemon so it can be killed on error
3289 file_read_firstline
($paths->{pid
}) =~ m/(\d+)/;
3293 sub vga_conf_has_spice
{
3296 my $vgaconf = parse_vga
($vga);
3297 my $vgatype = $vgaconf->{type
};
3298 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3305 return get_host_arch
() eq $arch;
3310 return $conf->{arch
} // get_host_arch
();
3313 my $default_machines = {
3318 sub get_installed_machine_version
{
3319 my ($kvmversion) = @_;
3320 $kvmversion = kvm_user_version
() if !defined($kvmversion);
3321 $kvmversion =~ m/^(\d+\.\d+)/;
3325 sub windows_get_pinned_machine_version
{
3326 my ($machine, $base_version, $kvmversion) = @_;
3328 my $pin_version = $base_version;
3329 if (!defined($base_version) ||
3330 !PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($base_version, $kvmversion)
3332 $pin_version = get_installed_machine_version
($kvmversion);
3334 if (!$machine || $machine eq 'pc') {
3335 $machine = "pc-i440fx-$pin_version";
3336 } elsif ($machine eq 'q35') {
3337 $machine = "pc-q35-$pin_version";
3338 } elsif ($machine eq 'virt') {
3339 $machine = "virt-$pin_version";
3341 warn "unknown machine type '$machine', not touching that!\n";
3347 sub get_vm_machine
{
3348 my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
3350 my $machine = $forcemachine || $conf->{machine
};
3352 if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
3353 $kvmversion //= kvm_user_version
();
3354 # we must pin Windows VMs without a specific version to 5.1, as 5.2 fixed a bug in ACPI
3355 # layout which confuses windows quite a bit and may result in various regressions..
3356 # see: https://lists.gnu.org/archive/html/qemu-devel/2021-02/msg08484.html
3357 if (windows_version
($conf->{ostype
})) {
3358 $machine = windows_get_pinned_machine_version
($machine, '5.1', $kvmversion);
3361 $machine ||= $default_machines->{$arch};
3362 if ($add_pve_version) {
3363 my $pvever = PVE
::QemuServer
::Machine
::get_pve_version
($kvmversion);
3364 $machine .= "+pve$pvever";
3368 if ($add_pve_version && $machine !~ m/\+pve\d+?(?:\.pxe)?$/) {
3369 my $is_pxe = $machine =~ m/^(.*?)\.pxe$/;
3370 $machine = $1 if $is_pxe;
3372 # for version-pinned machines that do not include a pve-version (e.g.
3373 # pc-q35-4.1), we assume 0 to keep them stable in case we bump
3374 $machine .= '+pve0';
3376 $machine .= '.pxe' if $is_pxe;
3382 sub get_ovmf_files
($$$) {
3383 my ($arch, $efidisk, $smm) = @_;
3385 my $types = $OVMF->{$arch}
3386 or die "no OVMF images known for architecture '$arch'\n";
3388 my $type = 'default';
3389 if (defined($efidisk->{efitype
}) && $efidisk->{efitype
} eq '4m') {
3390 $type = $smm ?
"4m" : "4m-no-smm";
3391 $type .= '-ms' if $efidisk->{'pre-enrolled-keys'};
3394 my ($ovmf_code, $ovmf_vars) = $types->{$type}->@*;
3395 die "EFI base image '$ovmf_code' not found\n" if ! -f
$ovmf_code;
3396 die "EFI vars image '$ovmf_vars' not found\n" if ! -f
$ovmf_vars;
3398 return ($ovmf_code, $ovmf_vars);
3402 aarch64
=> '/usr/bin/qemu-system-aarch64',
3403 x86_64
=> '/usr/bin/qemu-system-x86_64',
3405 sub get_command_for_arch
($) {
3407 return '/usr/bin/kvm' if is_native
($arch);
3409 my $cmd = $Arch2Qemu->{$arch}
3410 or die "don't know how to emulate architecture '$arch'\n";
3414 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
3415 # to use in a QEMU command line (-cpu element), first array_intersect the result
3416 # of query_supported_ with query_understood_. This is necessary because:
3418 # a) query_understood_ returns flags the host cannot use and
3419 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
3420 # flags, but CPU settings - with most of them being flags. Those settings
3421 # (and some flags, curiously) cannot be specified as a "-cpu" argument.
3423 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
3424 # expensive. If you need the value returned from this, you can get it much
3425 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
3426 # $accel being 'kvm' or 'tcg'.
3428 # pvestatd calls this function on startup and whenever the QEMU/KVM version
3429 # changes, automatically populating pmxcfs.
3431 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
3432 # since kvm and tcg machines support different flags
3434 sub query_supported_cpu_flags
{
3437 $arch //= get_host_arch
();
3438 my $default_machine = $default_machines->{$arch};
3442 # FIXME: Once this is merged, the code below should work for ARM as well:
3443 # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
3444 die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
3447 my $kvm_supported = defined(kvm_version
());
3448 my $qemu_cmd = get_command_for_arch
($arch);
3450 my $pidfile = PVE
::QemuServer
::Helpers
::pidfile_name
($fakevmid);
3452 # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
3453 my $query_supported_run_qemu = sub {
3459 '-machine', $default_machine,
3461 '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server=on,wait=off",
3462 '-mon', 'chardev=qmp,mode=control',
3463 '-pidfile', $pidfile,
3468 push @$cmd, '-accel', 'tcg';
3471 my $rc = run_command
($cmd, noerr
=> 1, quiet
=> 0);
3472 die "QEMU flag querying VM exited with code " . $rc if $rc;
3475 my $cmd_result = mon_cmd
(
3477 'query-cpu-model-expansion',
3479 model
=> { name
=> 'host' }
3482 my $props = $cmd_result->{model
}->{props
};
3483 foreach my $prop (keys %$props) {
3484 next if $props->{$prop} ne '1';
3485 # QEMU returns some flags multiple times, with '_', '.' or '-'
3486 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
3487 # We only keep those with underscores, to match /proc/cpuinfo
3488 $prop =~ s/\.|-/_/g;
3489 $flags->{$prop} = 1;
3494 # force stop with 10 sec timeout and 'nocheck', always stop, even if QMP failed
3495 vm_stop
(undef, $fakevmid, 1, 1, 10, 0, 1);
3499 return [ sort keys %$flags ];
3502 # We need to query QEMU twice, since KVM and TCG have different supported flags
3503 PVE
::QemuConfig-
>lock_config($fakevmid, sub {
3504 $flags->{tcg
} = eval { $query_supported_run_qemu->(0) };
3505 warn "warning: failed querying supported tcg flags: $@\n" if $@;
3507 if ($kvm_supported) {
3508 $flags->{kvm
} = eval { $query_supported_run_qemu->(1) };
3509 warn "warning: failed querying supported kvm flags: $@\n" if $@;
3516 # Understood CPU flags are written to a file at 'pve-qemu' compile time
3517 my $understood_cpu_flag_dir = "/usr/share/kvm";
3518 sub query_understood_cpu_flags
{
3519 my $arch = get_host_arch
();
3520 my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
3522 die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
3525 my $raw = file_get_contents
($filepath);
3526 $raw =~ s/^\s+|\s+$//g;
3527 my @flags = split(/\s+/, $raw);
3532 # Since commit 277d33454f77ec1d1e0bc04e37621e4dd2424b67 in pve-qemu, smm is not off by default
3533 # anymore. But smm=off seems to be required when using SeaBIOS and serial display.
3534 my sub should_disable_smm
{
3535 my ($conf, $vga) = @_;
3537 return (!defined($conf->{bios
}) || $conf->{bios
} eq 'seabios') &&
3538 $vga->{type
} && $vga->{type
} =~ m/^(serial\d+|none)$/;
3541 my sub print_ovmf_drive_commandlines
{
3542 my ($conf, $storecfg, $vmid, $arch, $q35, $version_guard) = @_;
3544 my $d = $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
3546 my ($ovmf_code, $ovmf_vars) = get_ovmf_files
($arch, $d, $q35);
3548 my $var_drive_str = "if=pflash,unit=1,id=drive-efidisk0";
3550 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($d->{file
}, 1);
3551 my ($path, $format) = $d->@{'file', 'format'};
3553 $path = PVE
::Storage
::path
($storecfg, $d->{file
});
3554 if (!defined($format)) {
3555 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
3556 $format = qemu_img_format
($scfg, $volname);
3558 } elsif (!defined($format)) {
3559 die "efidisk format must be specified\n";
3561 # SPI flash does lots of read-modify-write OPs, without writeback this gets really slow #3329
3562 if ($path =~ m/^rbd:/) {
3563 $var_drive_str .= ',cache=writeback';
3564 $path .= ':rbd_cache_policy=writeback'; # avoid write-around, we *need* to cache writes too
3566 $var_drive_str .= ",format=$format,file=$path";
3568 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $format eq 'raw' && $version_guard->(4, 1, 2);
3569 $var_drive_str .= ',readonly=on' if drive_is_read_only
($conf, $d);
3571 log_warn
("no efidisk configured! Using temporary efivars disk.");
3572 my $path = "/tmp/$vmid-ovmf.fd";
3573 PVE
::Tools
::file_copy
($ovmf_vars, $path, -s
$ovmf_vars);
3574 $var_drive_str .= ",format=raw,file=$path";
3575 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $version_guard->(4, 1, 2);
3578 return ("if=pflash,unit=0,format=raw,readonly=on,file=$ovmf_code", $var_drive_str);
3581 sub config_to_command
{
3582 my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu,
3585 my ($globalFlags, $machineFlags, $rtcFlags) = ([], [], []);
3588 my $ostype = $conf->{ostype
};
3589 my $winversion = windows_version
($ostype);
3590 my $kvm = $conf->{kvm
};
3591 my $nodename = nodename
();
3593 my $arch = get_vm_arch
($conf);
3594 my $kvm_binary = get_command_for_arch
($arch);
3595 my $kvmver = kvm_user_version
($kvm_binary);
3597 if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
3598 $kvmver //= "undefined";
3599 die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
3602 my $add_pve_version = min_version
($kvmver, 4, 1);
3604 my $machine_type = get_vm_machine
($conf, $forcemachine, $arch, $add_pve_version);
3605 my $machine_version = extract_version
($machine_type, $kvmver);
3606 $kvm //= 1 if is_native
($arch);
3608 $machine_version =~ m/(\d+)\.(\d+)/;
3609 my ($machine_major, $machine_minor) = ($1, $2);
3611 if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
3612 warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
3613 } elsif (!min_version
($kvmver, $machine_major, $machine_minor)) {
3614 die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type',"
3615 ." please upgrade node '$nodename'\n"
3616 } elsif (!PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($machine_version, $kvmver)) {
3617 my $max_pve_version = PVE
::QemuServer
::Machine
::get_pve_version
($machine_version);
3618 die "Installed qemu-server (max feature level for $machine_major.$machine_minor is"
3619 ." pve$max_pve_version) is too old to run machine type '$machine_type', please upgrade"
3620 ." node '$nodename'\n";
3623 # if a specific +pve version is required for a feature, use $version_guard
3624 # instead of min_version to allow machines to be run with the minimum
3626 my $required_pve_version = 0;
3627 my $version_guard = sub {
3628 my ($major, $minor, $pve) = @_;
3629 return 0 if !min_version
($machine_version, $major, $minor, $pve);
3630 my $max_pve = PVE
::QemuServer
::Machine
::get_pve_version
("$major.$minor");
3631 return 1 if min_version
($machine_version, $major, $minor, $max_pve+1);
3632 $required_pve_version = $pve if $pve && $pve > $required_pve_version;
3636 if ($kvm && !defined kvm_version
()) {
3637 die "KVM virtualisation configured, but not available. Either disable in VM configuration"
3638 ." or enable in BIOS.\n";
3641 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
3642 my $hotplug_features = parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
3643 my $use_old_bios_files = undef;
3644 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
3647 if ($conf->{affinity
}) {
3648 push @$cmd, '/usr/bin/taskset', '--cpu-list', '--all-tasks', $conf->{affinity
};
3651 push @$cmd, $kvm_binary;
3653 push @$cmd, '-id', $vmid;
3655 my $vmname = $conf->{name
} || "vm$vmid";
3657 push @$cmd, '-name', "$vmname,debug-threads=on";
3659 push @$cmd, '-no-shutdown';
3663 my $qmpsocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid);
3664 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server=on,wait=off";
3665 push @$cmd, '-mon', "chardev=qmp,mode=control";
3667 if (min_version
($machine_version, 2, 12)) {
3668 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3669 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3672 push @$cmd, '-pidfile' , PVE
::QemuServer
::Helpers
::pidfile_name
($vmid);
3674 push @$cmd, '-daemonize';
3676 if ($conf->{smbios1
}) {
3677 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
3678 if ($smbios_conf->{base64
}) {
3679 # Do not pass base64 flag to qemu
3680 delete $smbios_conf->{base64
};
3681 my $smbios_string = "";
3682 foreach my $key (keys %$smbios_conf) {
3684 if ($key eq "uuid") {
3685 $value = $smbios_conf->{uuid
}
3687 $value = decode_base64
($smbios_conf->{$key});
3689 # qemu accepts any binary data, only commas need escaping by double comma
3691 $smbios_string .= "," . $key . "=" . $value if $value;
3693 push @$cmd, '-smbios', "type=1" . $smbios_string;
3695 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3699 if ($conf->{bios
} && $conf->{bios
} eq 'ovmf') {
3700 my ($code_drive_str, $var_drive_str) =
3701 print_ovmf_drive_commandlines
($conf, $storecfg, $vmid, $arch, $q35, $version_guard);
3702 push $cmd->@*, '-drive', $code_drive_str;
3703 push $cmd->@*, '-drive', $var_drive_str;
3706 if ($q35) { # tell QEMU to load q35 config early
3707 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3708 if (min_version
($machine_version, 4, 0)) {
3709 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3711 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3715 if (defined(my $fixups = qemu_created_version_fixups
($conf, $forcemachine, $kvmver))) {
3716 push @$cmd, $fixups->@*;
3719 if ($conf->{vmgenid
}) {
3720 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid
};
3723 # add usb controllers
3724 my @usbcontrollers = PVE
::QemuServer
::USB
::get_usb_controllers
(
3725 $conf, $bridges, $arch, $machine_type, $usbdesc->{format
}, $MAX_USB_DEVICES, $machine_version);
3726 push @$devices, @usbcontrollers if @usbcontrollers;
3727 my $vga = parse_vga
($conf->{vga
});
3729 my $qxlnum = vga_conf_has_spice
($conf->{vga
});
3730 $vga->{type
} = 'qxl' if $qxlnum;
3732 if (!$vga->{type
}) {
3733 if ($arch eq 'aarch64') {
3734 $vga->{type
} = 'virtio';
3735 } elsif (min_version
($machine_version, 2, 9)) {
3736 $vga->{type
} = (!$winversion || $winversion >= 6) ?
'std' : 'cirrus';
3738 $vga->{type
} = ($winversion >= 6) ?
'std' : 'cirrus';
3742 # enable absolute mouse coordinates (needed by vnc)
3743 my $tablet = $conf->{tablet
};
3744 if (!defined($tablet)) {
3745 $tablet = $defaults->{tablet
};
3746 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3747 $tablet = 0 if $vga->{type
} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3751 push @$devices, '-device', print_tabletdevice_full
($conf, $arch) if $tablet;
3752 my $kbd = print_keyboarddevice_full
($conf, $arch);
3753 push @$devices, '-device', $kbd if defined($kbd);
3756 my $bootorder = device_bootorder
($conf);
3758 # host pci device passthrough
3759 my ($kvm_off, $gpu_passthrough, $legacy_igd) = PVE
::QemuServer
::PCI
::print_hostpci_devices
(
3760 $vmid, $conf, $devices, $vga, $winversion, $q35, $bridges, $arch, $machine_type, $bootorder);
3763 my $usb_dev_features = {};
3764 $usb_dev_features->{spice_usb3
} = 1 if min_version
($machine_version, 4, 0);
3766 my @usbdevices = PVE
::QemuServer
::USB
::get_usb_devices
(
3767 $conf, $usbdesc->{format
}, $MAX_USB_DEVICES, $usb_dev_features, $bootorder, $machine_version);
3768 push @$devices, @usbdevices if @usbdevices;
3771 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3772 my $path = $conf->{"serial$i"} or next;
3773 if ($path eq 'socket') {
3774 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3775 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server=on,wait=off";
3776 # On aarch64, serial0 is the UART device. QEMU only allows
3777 # connecting UART devices via the '-serial' command line, as
3778 # the device has a fixed slot on the hardware...
3779 if ($arch eq 'aarch64' && $i == 0) {
3780 push @$devices, '-serial', "chardev:serial$i";
3782 push @$devices, '-device', "isa-serial,chardev=serial$i";
3785 die "no such serial device\n" if ! -c
$path;
3786 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3787 push @$devices, '-device', "isa-serial,chardev=serial$i";
3792 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3793 if (my $path = $conf->{"parallel$i"}) {
3794 die "no such parallel device\n" if ! -c
$path;
3795 my $devtype = $path =~ m!^/dev/usb/lp! ?
'tty' : 'parport';
3796 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3797 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3801 if (min_version
($machine_version, 4, 0) && (my $audio = conf_has_audio
($conf))) {
3802 my $audiopciaddr = print_pci_addr
("audio0", $bridges, $arch, $machine_type);
3803 my $audio_devs = audio_devs
($audio, $audiopciaddr, $machine_version);
3804 push @$devices, @$audio_devs;
3807 add_tpm_device
($vmid, $devices, $conf);
3810 $sockets = $conf->{smp
} if $conf->{smp
}; # old style - no longer iused
3811 $sockets = $conf->{sockets
} if $conf->{sockets
};
3813 my $cores = $conf->{cores
} || 1;
3815 my $maxcpus = $sockets * $cores;
3817 my $vcpus = $conf->{vcpus
} ?
$conf->{vcpus
} : $maxcpus;
3819 my $allowed_vcpus = $cpuinfo->{cpus
};
3821 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n" if ($allowed_vcpus < $maxcpus);
3823 if ($hotplug_features->{cpu
} && min_version
($machine_version, 2, 7)) {
3824 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3825 for (my $i = 2; $i <= $vcpus; $i++) {
3826 my $cpustr = print_cpu_device
($conf,$i);
3827 push @$cmd, '-device', $cpustr;
3832 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3834 push @$cmd, '-nodefaults';
3836 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3838 push @$cmd, '-no-acpi' if defined($conf->{acpi
}) && $conf->{acpi
} == 0;
3840 push @$cmd, '-no-reboot' if defined($conf->{reboot
}) && $conf->{reboot
} == 0;
3842 if ($vga->{type
} && $vga->{type
} !~ m/^serial\d+$/ && $vga->{type
} ne 'none'){
3843 push @$devices, '-device', print_vga_device
(
3844 $conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3846 push @$cmd, '-display', 'egl-headless,gl=core' if $vga->{type
} eq 'virtio-gl'; # VIRGL
3848 my $socket = PVE
::QemuServer
::Helpers
::vnc_socket
($vmid);
3849 push @$cmd, '-vnc', "unix:$socket,password=on";
3851 push @$cmd, '-vga', 'none' if $vga->{type
} eq 'none';
3852 push @$cmd, '-nographic';
3856 my $tdf = defined($conf->{tdf
}) ?
$conf->{tdf
} : $defaults->{tdf
};
3857 my $useLocaltime = $conf->{localtime};
3859 if ($winversion >= 5) { # windows
3860 $useLocaltime = 1 if !defined($conf->{localtime});
3862 # use time drift fix when acpi is enabled
3863 if (!(defined($conf->{acpi
}) && $conf->{acpi
} == 0)) {
3864 $tdf = 1 if !defined($conf->{tdf
});
3868 if ($winversion >= 6) {
3869 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3870 push @$cmd, '-no-hpet';
3873 push @$rtcFlags, 'driftfix=slew' if $tdf;
3875 if ($conf->{startdate
} && $conf->{startdate
} ne 'now') {
3876 push @$rtcFlags, "base=$conf->{startdate}";
3877 } elsif ($useLocaltime) {
3878 push @$rtcFlags, 'base=localtime';
3882 push @$cmd, '-cpu', $forcecpu;
3884 push @$cmd, get_cpu_options
($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough);
3887 PVE
::QemuServer
::Memory
::config
($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3889 push @$cmd, '-S' if $conf->{freeze
};
3891 push @$cmd, '-k', $conf->{keyboard
} if defined($conf->{keyboard
});
3893 my $guest_agent = parse_guest_agent
($conf);
3895 if ($guest_agent->{enabled
}) {
3896 my $qgasocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid, 1);
3897 push @$devices, '-chardev', "socket,path=$qgasocket,server=on,wait=off,id=qga0";
3899 if (!$guest_agent->{type
} || $guest_agent->{type
} eq 'virtio') {
3900 my $pciaddr = print_pci_addr
("qga0", $bridges, $arch, $machine_type);
3901 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3902 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3903 } elsif ($guest_agent->{type
} eq 'isa') {
3904 push @$devices, '-device', "isa-serial,chardev=qga0";
3908 my $rng = $conf->{rng0
} ? parse_rng
($conf->{rng0
}) : undef;
3909 if ($rng && $version_guard->(4, 1, 2)) {
3910 check_rng_source
($rng->{source
});
3912 my $max_bytes = $rng->{max_bytes
} // $rng_fmt->{max_bytes
}->{default};
3913 my $period = $rng->{period
} // $rng_fmt->{period
}->{default};
3914 my $limiter_str = "";
3916 $limiter_str = ",max-bytes=$max_bytes,period=$period";
3919 my $rng_addr = print_pci_addr
("rng0", $bridges, $arch, $machine_type);
3920 push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3921 push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3926 if ($qxlnum || $vga->{type
} =~ /^virtio/) {
3929 for (my $i = 1; $i < $qxlnum; $i++){
3930 push @$devices, '-device', print_vga_device
(
3931 $conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3934 # assume other OS works like Linux
3935 my ($ram, $vram) = ("134217728", "67108864");
3936 if ($vga->{memory
}) {
3937 $ram = PVE
::Tools
::convert_size
($qxlnum*4*$vga->{memory
}, 'mb' => 'b');
3938 $vram = PVE
::Tools
::convert_size
($qxlnum*2*$vga->{memory
}, 'mb' => 'b');
3940 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3941 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3945 my $pciaddr = print_pci_addr
("spice", $bridges, $arch, $machine_type);
3947 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
3948 my @nodeaddrs = PVE
::Tools
::getaddrinfo_all
('localhost', family
=> $pfamily);
3949 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3951 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3952 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3953 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3955 my $localhost = PVE
::Network
::addr_to_ip
($nodeaddrs[0]->{addr
});
3956 $spice_port = PVE
::Tools
::next_spice_port
($pfamily, $localhost);
3958 my $spice_enhancement_str = $conf->{spice_enhancements
} // '';
3959 my $spice_enhancement = parse_property_string
($spice_enhancements_fmt, $spice_enhancement_str);
3960 if ($spice_enhancement->{foldersharing
}) {
3961 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3962 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3965 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3966 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}"
3967 if $spice_enhancement->{videostreaming
};
3969 push @$devices, '-spice', "$spice_opts";
3972 # enable balloon by default, unless explicitly disabled
3973 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
3974 my $pciaddr = print_pci_addr
("balloon0", $bridges, $arch, $machine_type);
3975 my $ballooncmd = "virtio-balloon-pci,id=balloon0$pciaddr";
3976 $ballooncmd .= ",free-page-reporting=on" if min_version
($machine_version, 6, 2);
3977 push @$devices, '-device', $ballooncmd;
3980 if ($conf->{watchdog
}) {
3981 my $wdopts = parse_watchdog
($conf->{watchdog
});
3982 my $pciaddr = print_pci_addr
("watchdog", $bridges, $arch, $machine_type);
3983 my $watchdog = $wdopts->{model
} || 'i6300esb';
3984 push @$devices, '-device', "$watchdog$pciaddr";
3985 push @$devices, '-watchdog-action', $wdopts->{action
} if $wdopts->{action
};
3989 my $scsicontroller = {};
3990 my $ahcicontroller = {};
3991 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : $defaults->{scsihw
};
3993 # Add iscsi initiator name if available
3994 if (my $initiator = get_initiator_name
()) {
3995 push @$devices, '-iscsi', "initiator-name=$initiator";
3998 PVE
::QemuConfig-
>foreach_volume($conf, sub {
3999 my ($ds, $drive) = @_;
4001 if (PVE
::Storage
::parse_volume_id
($drive->{file
}, 1)) {
4002 check_volume_storage_type
($storecfg, $drive->{file
});
4003 push @$vollist, $drive->{file
};
4006 # ignore efidisk here, already added in bios/fw handling code above
4007 return if $drive->{interface
} eq 'efidisk';
4009 return if $drive->{interface
} eq 'tpmstate';
4011 $use_virtio = 1 if $ds =~ m/^virtio/;
4013 $drive->{bootindex
} = $bootorder->{$ds} if $bootorder->{$ds};
4015 if ($drive->{interface
} eq 'virtio'){
4016 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread
};
4019 if ($drive->{interface
} eq 'scsi') {
4021 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
4023 die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
4024 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
4026 my $pciaddr = print_pci_addr
("$controller_prefix$controller", $bridges, $arch, $machine_type);
4027 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ?
"virtio-scsi-pci" : $scsihw;
4030 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{iothread
}){
4031 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4032 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4033 } elsif ($drive->{iothread
}) {
4035 "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n"
4040 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{queues
}){
4041 $queues = ",num_queues=$drive->{queues}";
4044 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues"
4045 if !$scsicontroller->{$controller};
4046 $scsicontroller->{$controller}=1;
4049 if ($drive->{interface
} eq 'sata') {
4050 my $controller = int($drive->{index} / $PVE::QemuServer
::Drive
::MAX_SATA_DISKS
);
4051 my $pciaddr = print_pci_addr
("ahci$controller", $bridges, $arch, $machine_type);
4052 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr"
4053 if !$ahcicontroller->{$controller};
4054 $ahcicontroller->{$controller}=1;
4057 my $pbs_conf = $pbs_backing->{$ds};
4058 my $pbs_name = undef;
4060 $pbs_name = "drive-$ds-pbs";
4061 push @$devices, '-blockdev', print_pbs_blockdev
($pbs_conf, $pbs_name);
4064 my $drive_cmd = print_drive_commandline_full
(
4065 $storecfg, $vmid, $drive, $pbs_name, min_version
($kvmver, 6, 0));
4067 # extra protection for templates, but SATA and IDE don't support it..
4068 $drive_cmd .= ',readonly=on' if drive_is_read_only
($conf, $drive);
4070 push @$devices, '-drive',$drive_cmd;
4071 push @$devices, '-device', print_drivedevice_full
(
4072 $storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4075 for (my $i = 0; $i < $MAX_NETS; $i++) {
4076 my $netname = "net$i";
4078 next if !$conf->{$netname};
4079 my $d = parse_net
($conf->{$netname});
4081 # save the MAC addr here (could be auto-gen. in some odd setups) for FDB registering later?
4083 $use_virtio = 1 if $d->{model
} eq 'virtio';
4085 $d->{bootindex
} = $bootorder->{$netname} if $bootorder->{$netname};
4087 my $netdevfull = print_netdev_full
($vmid, $conf, $arch, $d, $netname);
4088 push @$devices, '-netdev', $netdevfull;
4090 my $netdevicefull = print_netdevice_full
(
4091 $vmid, $conf, $d, $netname, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version);
4093 push @$devices, '-device', $netdevicefull;
4096 if ($conf->{ivshmem
}) {
4097 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
4101 $bus = print_pcie_addr
("ivshmem");
4103 $bus = print_pci_addr
("ivshmem", $bridges, $arch, $machine_type);
4106 my $ivshmem_name = $ivshmem->{name
} // $vmid;
4107 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4109 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4110 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path"
4111 .",size=$ivshmem->{size}M";
4114 # pci.4 is nested in pci.1
4115 $bridges->{1} = 1 if $bridges->{4};
4117 if (!$q35) { # add pci bridges
4118 if (min_version
($machine_version, 2, 3)) {
4122 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4125 for my $k (sort {$b cmp $a} keys %$bridges) {
4126 next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
4129 if ($k == 2 && $legacy_igd) {
4132 my $pciaddr = print_pci_addr
("pci.$k_name", undef, $arch, $machine_type);
4133 my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
4135 if ($q35) { # add after -readconfig pve-q35.cfg
4136 splice @$devices, 2, 0, '-device', $devstr;
4138 unshift @$devices, '-device', $devstr if $k > 0;
4143 push @$machineFlags, 'accel=tcg';
4146 push @$machineFlags, 'smm=off' if should_disable_smm
($conf, $vga);
4148 my $machine_type_min = $machine_type;
4149 if ($add_pve_version) {
4150 $machine_type_min =~ s/\+pve\d+$//;
4151 $machine_type_min .= "+pve$required_pve_version";
4153 push @$machineFlags, "type=${machine_type_min}";
4155 push @$cmd, @$devices;
4156 push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
4157 push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
4158 push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
4160 if (my $vmstate = $conf->{vmstate
}) {
4161 my $statepath = PVE
::Storage
::path
($storecfg, $vmstate);
4162 push @$vollist, $vmstate;
4163 push @$cmd, '-loadstate', $statepath;
4164 print "activating and using '$vmstate' as vmstate\n";
4167 if (PVE
::QemuConfig-
>is_template($conf)) {
4168 # needed to workaround base volumes being read-only
4169 push @$cmd, '-snapshot';
4173 if ($conf->{args
}) {
4174 my $aa = PVE
::Tools
::split_args
($conf->{args
});
4178 return wantarray ?
($cmd, $vollist, $spice_port) : $cmd;
4181 sub check_rng_source
{
4184 # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
4185 die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
4188 my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
4189 if ($source eq '/dev/hwrng' && file_read_firstline
($rng_current) eq 'none') {
4190 # Needs to abort, otherwise QEMU crashes on first rng access. Note that rng_current cannot
4191 # be changed to 'none' manually, so once the VM is past this point, it's no longer an issue.
4192 die "Cannot start VM with passed-through RNG device: '/dev/hwrng' exists, but"
4193 ." '$rng_current' is set to 'none'. Ensure that a compatible hardware-RNG is attached"
4201 my $res = mon_cmd
($vmid, 'query-spice');
4203 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4206 sub vm_devices_list
{
4209 my $res = mon_cmd
($vmid, 'query-pci');
4210 my $devices_to_check = [];
4212 foreach my $pcibus (@$res) {
4213 push @$devices_to_check, @{$pcibus->{devices
}},
4216 while (@$devices_to_check) {
4218 for my $d (@$devices_to_check) {
4219 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4220 next if !$d->{'pci_bridge'} || !$d->{'pci_bridge'}->{devices
};
4222 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices
}});
4223 push @$to_check, @{$d->{'pci_bridge'}->{devices
}};
4225 $devices_to_check = $to_check;
4228 my $resblock = mon_cmd
($vmid, 'query-block');
4229 foreach my $block (@$resblock) {
4230 if($block->{device
} =~ m/^drive-(\S+)/){
4235 my $resmice = mon_cmd
($vmid, 'query-mice');
4236 foreach my $mice (@$resmice) {
4237 if ($mice->{name
} eq 'QEMU HID Tablet') {
4238 $devices->{tablet
} = 1;
4243 # for usb devices there is no query-usb
4244 # but we can iterate over the entries in
4245 # qom-list path=/machine/peripheral
4246 my $resperipheral = mon_cmd
($vmid, 'qom-list', path
=> '/machine/peripheral');
4247 foreach my $per (@$resperipheral) {
4248 if ($per->{name
} =~ m/^usb(?:redirdev)?\d+$/) {
4249 $devices->{$per->{name
}} = 1;
4257 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4259 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
4261 my $devices_list = vm_devices_list
($vmid);
4262 return 1 if defined($devices_list->{$deviceid});
4264 # add PCI bridge if we need it for the device
4265 qemu_add_pci_bridge
($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type);
4267 if ($deviceid eq 'tablet') {
4268 qemu_deviceadd
($vmid, print_tabletdevice_full
($conf, $arch));
4269 } elsif ($deviceid eq 'keyboard') {
4270 qemu_deviceadd
($vmid, print_keyboarddevice_full
($conf, $arch));
4271 } elsif ($deviceid =~ m/^usbredirdev(\d+)$/) {
4273 qemu_spice_usbredir_chardev_add
($vmid, "usbredirchardev$id");
4274 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_spice_usbdevice
($id, "xhci", $id + 1));
4275 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4276 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_usbdevice_full
($conf, $deviceid, $device, {}, $1 + 1));
4277 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4278 qemu_iothread_add
($vmid, $deviceid, $device);
4280 qemu_driveadd
($storecfg, $vmid, $device);
4281 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4283 qemu_deviceadd
($vmid, $devicefull);
4284 eval { qemu_deviceaddverify
($vmid, $deviceid); };
4286 eval { qemu_drivedel
($vmid, $deviceid); };
4290 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4291 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : "lsi";
4292 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4293 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ?
"virtio-scsi-pci" : $scsihw;
4295 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4297 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread
}) {
4298 qemu_iothread_add
($vmid, $deviceid, $device);
4299 $devicefull .= ",iothread=iothread-$deviceid";
4302 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues
}) {
4303 $devicefull .= ",num_queues=$device->{queues}";
4306 qemu_deviceadd
($vmid, $devicefull);
4307 qemu_deviceaddverify
($vmid, $deviceid);
4308 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4309 qemu_findorcreatescsihw
($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4310 qemu_driveadd
($storecfg, $vmid, $device);
4312 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4313 eval { qemu_deviceadd
($vmid, $devicefull); };
4315 eval { qemu_drivedel
($vmid, $deviceid); };
4319 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4320 return if !qemu_netdevadd
($vmid, $conf, $arch, $device, $deviceid);
4322 my $machine_type = PVE
::QemuServer
::Machine
::qemu_machine_pxe
($vmid, $conf);
4323 my $machine_version = PVE
::QemuServer
::Machine
::extract_version
($machine_type);
4324 my $use_old_bios_files = undef;
4325 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
4327 my $netdevicefull = print_netdevice_full
(
4328 $vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type, $machine_version);
4329 qemu_deviceadd
($vmid, $netdevicefull);
4331 qemu_deviceaddverify
($vmid, $deviceid);
4332 qemu_set_link_status
($vmid, $deviceid, !$device->{link_down
});
4335 eval { qemu_netdevdel
($vmid, $deviceid); };
4339 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4341 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4342 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4344 qemu_deviceadd
($vmid, $devicefull);
4345 qemu_deviceaddverify
($vmid, $deviceid);
4347 die "can't hotplug device '$deviceid'\n";
4353 # fixme: this should raise exceptions on error!
4354 sub vm_deviceunplug
{
4355 my ($vmid, $conf, $deviceid) = @_;
4357 my $devices_list = vm_devices_list
($vmid);
4358 return 1 if !defined($devices_list->{$deviceid});
4360 my $bootdisks = PVE
::QemuServer
::Drive
::get_bootdisks
($conf);
4361 die "can't unplug bootdisk '$deviceid'\n" if grep {$_ eq $deviceid} @$bootdisks;
4363 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard' || $deviceid eq 'xhci') {
4364 qemu_devicedel
($vmid, $deviceid);
4365 } elsif ($deviceid =~ m/^usbredirdev\d+$/) {
4366 qemu_devicedel
($vmid, $deviceid);
4367 qemu_devicedelverify
($vmid, $deviceid);
4368 } elsif ($deviceid =~ m/^usb\d+$/) {
4369 qemu_devicedel
($vmid, $deviceid);
4370 qemu_devicedelverify
($vmid, $deviceid);
4371 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4372 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4374 qemu_devicedel
($vmid, $deviceid);
4375 qemu_devicedelverify
($vmid, $deviceid);
4376 qemu_drivedel
($vmid, $deviceid);
4377 qemu_iothread_del
($vmid, $deviceid, $device);
4378 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4379 qemu_devicedel
($vmid, $deviceid);
4380 qemu_devicedelverify
($vmid, $deviceid);
4381 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4382 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4384 qemu_devicedel
($vmid, $deviceid);
4385 qemu_devicedelverify
($vmid, $deviceid);
4386 qemu_drivedel
($vmid, $deviceid);
4387 qemu_deletescsihw
($conf, $vmid, $deviceid);
4389 qemu_iothread_del
($vmid, "virtioscsi$device->{index}", $device)
4390 if $conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single');
4391 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4392 qemu_devicedel
($vmid, $deviceid);
4393 qemu_devicedelverify
($vmid, $deviceid);
4394 qemu_netdevdel
($vmid, $deviceid);
4396 die "can't unplug device '$deviceid'\n";
4402 sub qemu_spice_usbredir_chardev_add
{
4403 my ($vmid, $id) = @_;
4405 mon_cmd
($vmid, "chardev-add" , (
4416 sub qemu_deviceadd
{
4417 my ($vmid, $devicefull) = @_;
4419 $devicefull = "driver=".$devicefull;
4420 my %options = split(/[=,]/, $devicefull);
4422 mon_cmd
($vmid, "device_add" , %options);
4425 sub qemu_devicedel
{
4426 my ($vmid, $deviceid) = @_;
4428 my $ret = mon_cmd
($vmid, "device_del", id
=> $deviceid);
4431 sub qemu_iothread_add
{
4432 my ($vmid, $deviceid, $device) = @_;
4434 if ($device->{iothread
}) {
4435 my $iothreads = vm_iothreads_list
($vmid);
4436 qemu_objectadd
($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4440 sub qemu_iothread_del
{
4441 my ($vmid, $deviceid, $device) = @_;
4443 if ($device->{iothread
}) {
4444 my $iothreads = vm_iothreads_list
($vmid);
4445 qemu_objectdel
($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4449 sub qemu_objectadd
{
4450 my ($vmid, $objectid, $qomtype) = @_;
4452 mon_cmd
($vmid, "object-add", id
=> $objectid, "qom-type" => $qomtype);
4457 sub qemu_objectdel
{
4458 my ($vmid, $objectid) = @_;
4460 mon_cmd
($vmid, "object-del", id
=> $objectid);
4466 my ($storecfg, $vmid, $device) = @_;
4468 my $kvmver = get_running_qemu_version
($vmid);
4469 my $io_uring = min_version
($kvmver, 6, 0);
4470 my $drive = print_drive_commandline_full
($storecfg, $vmid, $device, undef, $io_uring);
4471 $drive =~ s/\\/\\\\/g;
4472 my $ret = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "drive_add auto \"$drive\"");
4474 # If the command succeeds qemu prints: "OK
"
4475 return 1 if $ret =~ m/OK/s;
4477 die "adding drive failed
: $ret\n";
4481 my ($vmid, $deviceid) = @_;
4483 my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-
$deviceid");
4486 return 1 if $ret eq "";
4488 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4489 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4491 die "deleting drive
$deviceid failed
: $ret\n";
4494 sub qemu_deviceaddverify {
4495 my ($vmid, $deviceid) = @_;
4497 for (my $i = 0; $i <= 5; $i++) {
4498 my $devices_list = vm_devices_list($vmid);
4499 return 1 if defined($devices_list->{$deviceid});
4503 die "error on hotplug device
'$deviceid'\n";
4507 sub qemu_devicedelverify {
4508 my ($vmid, $deviceid) = @_;
4510 # need to verify that the device is correctly removed as device_del
4511 # is async and empty return is not reliable
4513 for (my $i = 0; $i <= 5; $i++) {
4514 my $devices_list = vm_devices_list($vmid);
4515 return 1 if !defined($devices_list->{$deviceid});
4519 die "error on hot-unplugging device
'$deviceid'\n";
4522 sub qemu_findorcreatescsihw {
4523 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4525 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4527 my $scsihwid="$controller_prefix$controller";
4528 my $devices_list = vm_devices_list($vmid);
4530 if (!defined($devices_list->{$scsihwid})) {
4531 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4537 sub qemu_deletescsihw {
4538 my ($conf, $vmid, $opt) = @_;
4540 my $device = parse_drive($opt, $conf->{$opt});
4542 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4543 vm_deviceunplug($vmid, $conf, "virtioscsi
$device->{index}");
4547 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4549 my $devices_list = vm_devices_list($vmid);
4550 foreach my $opt (keys %{$devices_list}) {
4551 if (is_valid_drivename($opt)) {
4552 my $drive = parse_drive($opt, $conf->{$opt});
4553 if ($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4559 my $scsihwid="scsihw
$controller";
4561 vm_deviceunplug($vmid, $conf, $scsihwid);
4566 sub qemu_add_pci_bridge {
4567 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4573 print_pci_addr($device, $bridges, $arch, $machine_type);
4575 while (my ($k, $v) = each %$bridges) {
4578 return 1 if !defined($bridgeid) || $bridgeid < 1;
4580 my $bridge = "pci
.$bridgeid";
4581 my $devices_list = vm_devices_list($vmid);
4583 if (!defined($devices_list->{$bridge})) {
4584 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4590 sub qemu_set_link_status {
4591 my ($vmid, $device, $up) = @_;
4593 mon_cmd($vmid, "set_link
", name => $device,
4594 up => $up ? JSON::true : JSON::false);
4597 sub qemu_netdevadd {
4598 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4600 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4601 my %options = split(/[=,]/, $netdev);
4603 if (defined(my $vhost = $options{vhost})) {
4604 $options{vhost} = JSON::boolean(PVE::JSONSchema::parse_boolean($vhost));
4607 if (defined(my $queues = $options{queues})) {
4608 $options{queues} = $queues + 0;
4611 mon_cmd($vmid, "netdev_add
", %options);
4615 sub qemu_netdevdel {
4616 my ($vmid, $deviceid) = @_;
4618 mon_cmd($vmid, "netdev_del
", id => $deviceid);
4621 sub qemu_usb_hotplug {
4622 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4626 # remove the old one first
4627 vm_deviceunplug($vmid, $conf, $deviceid);
4629 # check if xhci controller is necessary and available
4630 my $devicelist = vm_devices_list($vmid);
4632 if (!$devicelist->{xhci}) {
4633 my $pciaddr = print_pci_addr("xhci
", undef, $arch, $machine_type);
4634 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_qemu_xhci_controller($pciaddr));
4637 # print_usbdevice_full expects the parsed device
4638 my $d = parse_usb_device($device->{host});
4639 $d->{usb3} = $device->{usb3};
4642 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4645 sub qemu_cpu_hotplug {
4646 my ($vmid, $conf, $vcpus) = @_;
4648 my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4651 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4652 $sockets = $conf->{sockets} if $conf->{sockets};
4653 my $cores = $conf->{cores} || 1;
4654 my $maxcpus = $sockets * $cores;
4656 $vcpus = $maxcpus if !$vcpus;
4658 die "you can
't add more vcpus than maxcpus\n"
4659 if $vcpus > $maxcpus;
4661 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4663 if ($vcpus < $currentvcpus) {
4665 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4667 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4668 qemu_devicedel($vmid, "cpu$i");
4670 my $currentrunningvcpus = undef;
4672 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4673 last if scalar(@{$currentrunningvcpus}) == $i-1;
4674 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4678 #update conf after each succesfull cpu unplug
4679 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4680 PVE::QemuConfig->write_config($vmid, $conf);
4683 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4689 my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4690 die "vcpus in running vm does not match its configuration\n"
4691 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4693 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4695 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4696 my $cpustr = print_cpu_device($conf, $i);
4697 qemu_deviceadd($vmid, $cpustr);
4700 my $currentrunningvcpus = undef;
4702 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4703 last if scalar(@{$currentrunningvcpus}) == $i;
4704 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4708 #update conf after each succesfull cpu hotplug
4709 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4710 PVE::QemuConfig->write_config($vmid, $conf);
4714 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4715 mon_cmd($vmid, "cpu-add", id => int($i));
4720 sub qemu_block_set_io_throttle {
4721 my ($vmid, $deviceid,
4722 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4723 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4724 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4725 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4727 return if !check_running($vmid) ;
4729 mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4731 bps_rd => int($bps_rd),
4732 bps_wr => int($bps_wr),
4734 iops_rd => int($iops_rd),
4735 iops_wr => int($iops_wr),
4736 bps_max => int($bps_max),
4737 bps_rd_max => int($bps_rd_max),
4738 bps_wr_max => int($bps_wr_max),
4739 iops_max => int($iops_max),
4740 iops_rd_max => int($iops_rd_max),
4741 iops_wr_max => int($iops_wr_max),
4742 bps_max_length => int($bps_max_length),
4743 bps_rd_max_length => int($bps_rd_max_length),
4744 bps_wr_max_length => int($bps_wr_max_length),
4745 iops_max_length => int($iops_max_length),
4746 iops_rd_max_length => int($iops_rd_max_length),
4747 iops_wr_max_length => int($iops_wr_max_length),
4752 sub qemu_block_resize {
4753 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4755 my $running = check_running($vmid);
4757 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4759 return if !$running;
4761 my $padding = (1024 - $size % 1024) % 1024;
4762 $size = $size + $padding;
4767 device => $deviceid,
4773 sub qemu_volume_snapshot {
4774 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4776 my $running = check_running($vmid);
4778 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4779 mon_cmd($vmid, 'blockdev-snapshot-internal-sync
', device => $deviceid, name => $snap);
4781 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4785 sub qemu_volume_snapshot_delete {
4786 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4788 my $running = check_running($vmid);
4793 my $conf = PVE::QemuConfig->load_config($vmid);
4794 PVE::QemuConfig->foreach_volume($conf, sub {
4795 my ($ds, $drive) = @_;
4796 $running = 1 if $drive->{file} eq $volid;
4800 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4801 mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync
', device => $deviceid, name => $snap);
4803 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4807 sub set_migration_caps {
4808 my ($vmid, $savevm) = @_;
4810 my $qemu_support = eval { mon_cmd($vmid, "query-proxmox-support") };
4812 my $bitmap_prop = $savevm ? 'pbs-dirty-bitmap-savevm
' : 'pbs-dirty-bitmap-migration
';
4813 my $dirty_bitmaps = $qemu_support->{$bitmap_prop} ? 1 : 0;
4818 "auto-converge" => 1,
4820 "x-rdma-pin-all" => 0,
4823 "dirty-bitmaps" => $dirty_bitmaps,
4826 my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4828 for my $supported_capability (@$supported_capabilities) {
4830 capability => $supported_capability->{capability},
4831 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4835 mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4839 my ($conf, $func, @param) = @_;
4843 my $test_volid = sub {
4844 my ($key, $drive, $snapname) = @_;
4846 my $volid = $drive->{file};
4849 $volhash->{$volid}->{cdrom} //= 1;
4850 $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4852 my $replicate = $drive->{replicate} // 1;
4853 $volhash->{$volid}->{replicate} //= 0;
4854 $volhash->{$volid}->{replicate} = 1 if $replicate;
4856 $volhash->{$volid}->{shared} //= 0;
4857 $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4859 $volhash->{$volid}->{referenced_in_config} //= 0;
4860 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
4862 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4863 if defined($snapname);
4865 my $size = $drive->{size};
4866 $volhash->{$volid}->{size} //= $size if $size;
4868 $volhash->{$volid}->{is_vmstate} //= 0;
4869 $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate
';
4871 $volhash->{$volid}->{is_tpmstate} //= 0;
4872 $volhash->{$volid}->{is_tpmstate} = 1 if $key eq 'tpmstate0
';
4874 $volhash->{$volid}->{is_unused} //= 0;
4875 $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4877 $volhash->{$volid}->{drivename} = $key if is_valid_drivename($key);
4880 my $include_opts = {
4881 extra_keys => ['vmstate
'],
4882 include_unused => 1,
4885 PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4886 foreach my $snapname (keys %{$conf->{snapshots}}) {
4887 my $snap = $conf->{snapshots}->{$snapname};
4888 PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4891 foreach my $volid (keys %$volhash) {
4892 &$func($volid, $volhash->{$volid}, @param);
4896 my $fast_plug_option = {
4904 'vmstatestorage
' => 1,
4909 for my $opt (keys %$confdesc_cloudinit) {
4910 $fast_plug_option->{$opt} = 1;
4913 # hotplug changes in [PENDING]
4914 # $selection hash can be used to only apply specified options, for
4915 # example: { cores => 1 } (only apply changed 'cores
')
4916 # $errors ref is used to return error messages
4917 sub vmconfig_hotplug_pending {
4918 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4920 my $defaults = load_defaults();
4921 my $arch = get_vm_arch($conf);
4922 my $machine_type = get_vm_machine($conf, undef, $arch);
4924 # commit values which do not have any impact on running VM first
4925 # Note: those option cannot raise errors, we we do not care about
4926 # $selection and always apply them.
4928 my $add_error = sub {
4929 my ($opt, $msg) = @_;
4930 $errors->{$opt} = "hotplug problem - $msg";
4933 my $cloudinit_pending_properties = PVE::QemuServer::cloudinit_pending_properties();
4935 my $cloudinit_record_changed = sub {
4936 my ($conf, $opt, $old, $new) = @_;
4937 return if !$cloudinit_pending_properties->{$opt};
4939 my $ci = ($conf->{cloudinit} //= {});
4941 my $recorded = $ci->{$opt};
4942 my %added = map { $_ => 1 } PVE::Tools::split_list(delete($ci->{added}) // '');
4944 if (defined($new)) {
4945 if (defined($old)) {
4946 # an existing value is being modified
4947 if (defined($recorded)) {
4948 # the value was already not in sync
4949 if ($new eq $recorded) {
4950 # a value is being reverted to the cloud-init state:
4952 delete $added{$opt};
4954 # the value was changed multiple times, do nothing
4956 } elsif ($added{$opt}) {
4957 # the value had been marked as added and is being changed, do nothing
4959 # the value is new, record it:
4963 # a new value is being added
4964 if (defined($recorded)) {
4965 # it was already not in sync
4966 if ($new eq $recorded) {
4967 # a value is being reverted to the cloud-init state:
4969 delete $added{$opt};
4971 # the value had temporarily been removed, do nothing
4973 } elsif ($added{$opt}) {
4974 # the value had been marked as added already, do nothing
4976 # the value is new, add it
4980 } elsif (!defined($old)) {
4981 # a non-existent value is being removed? ignore...
4983 # a value is being deleted
4984 if (defined($recorded)) {
4985 # a value was already recorded, just keep it
4986 } elsif ($added{$opt}) {
4987 # the value was marked as added, remove it
4988 delete $added{$opt};
4990 # a previously unrecorded value is being removed, record the old value:
4995 my $added = join(',', sort keys %added);
4996 $ci->{added} = $added if length($added);
5000 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5001 if ($fast_plug_option->{$opt}) {
5002 my $new = delete $conf->{pending}->{$opt};
5003 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $new);
5004 $conf->{$opt} = $new;
5010 PVE::QemuConfig->write_config($vmid, $conf);
5013 my $ostype = $conf->{ostype};
5014 my $version = extract_version($machine_type, get_running_qemu_version($vmid));
5015 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
5016 my $usb_hotplug = $hotplug_features->{usb}
5017 && min_version($version, 7, 1)
5018 && defined($ostype) && ($ostype eq 'l26
' || windows_version($ostype) > 7);
5020 my $cgroup = PVE::QemuServer::CGroup->new($vmid);
5021 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
5023 foreach my $opt (sort keys %$pending_delete_hash) {
5024 next if $selection && !$selection->{$opt};
5025 my $force = $pending_delete_hash->{$opt}->{force};
5027 if ($opt eq 'hotplug
') {
5028 die "skip\n" if ($conf->{hotplug} =~ /memory/);
5029 } elsif ($opt eq 'tablet
') {
5030 die "skip\n" if !$hotplug_features->{usb};
5031 if ($defaults->{tablet}) {
5032 vm_deviceplug($storecfg, $conf, $vmid, 'tablet
', $arch, $machine_type);
5033 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard
', $arch, $machine_type)
5034 if $arch eq 'aarch64
';
5036 vm_deviceunplug($vmid, $conf, 'tablet
');
5037 vm_deviceunplug($vmid, $conf, 'keyboard
') if $arch eq 'aarch64
';
5039 } elsif ($opt =~ m/^usb(\d+)$/) {
5041 die "skip\n" if !$usb_hotplug;
5042 vm_deviceunplug($vmid, $conf, "usbredirdev$index"); # if it's a spice port
5043 vm_deviceunplug
($vmid, $conf, $opt);
5044 } elsif ($opt eq 'vcpus') {
5045 die "skip\n" if !$hotplug_features->{cpu
};
5046 qemu_cpu_hotplug
($vmid, $conf, undef);
5047 } elsif ($opt eq 'balloon') {
5048 # enable balloon device is not hotpluggable
5049 die "skip\n" if defined($conf->{balloon
}) && $conf->{balloon
} == 0;
5050 # here we reset the ballooning value to memory
5051 my $balloon = $conf->{memory
} || $defaults->{memory
};
5052 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5053 } elsif ($fast_plug_option->{$opt}) {
5055 } elsif ($opt =~ m/^net(\d+)$/) {
5056 die "skip\n" if !$hotplug_features->{network
};
5057 vm_deviceunplug
($vmid, $conf, $opt);
5058 } elsif (is_valid_drivename
($opt)) {
5059 die "skip\n" if !$hotplug_features->{disk
} || $opt =~ m/(ide|sata)(\d+)/;
5060 vm_deviceunplug
($vmid, $conf, $opt);
5061 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5062 } elsif ($opt =~ m/^memory$/) {
5063 die "skip\n" if !$hotplug_features->{memory
};
5064 PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults);
5065 } elsif ($opt eq 'cpuunits') {
5066 $cgroup->change_cpu_shares(undef);
5067 } elsif ($opt eq 'cpulimit') {
5068 $cgroup->change_cpu_quota(undef, undef); # reset, cgroup module can better decide values
5074 &$add_error($opt, $err) if $err ne "skip\n";
5076 my $old = delete $conf->{$opt};
5077 $cloudinit_record_changed->($conf, $opt, $old, undef);
5078 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5083 foreach my $opt (keys %{$conf->{pending
}}) {
5084 next if $selection && !$selection->{$opt};
5085 my $value = $conf->{pending
}->{$opt};
5087 if ($opt eq 'hotplug') {
5088 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug
} =~ /memory/);
5089 } elsif ($opt eq 'tablet') {
5090 die "skip\n" if !$hotplug_features->{usb
};
5092 vm_deviceplug
($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5093 vm_deviceplug
($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5094 if $arch eq 'aarch64';
5095 } elsif ($value == 0) {
5096 vm_deviceunplug
($vmid, $conf, 'tablet');
5097 vm_deviceunplug
($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5099 } elsif ($opt =~ m/^usb(\d+)$/) {
5101 die "skip\n" if !$usb_hotplug;
5102 my $d = eval { parse_property_string
($usbdesc->{format
}, $value) };
5104 if ($d->{host
} eq 'spice') {
5105 $id = "usbredirdev$index";
5107 qemu_usb_hotplug
($storecfg, $conf, $vmid, $id, $d, $arch, $machine_type);
5108 } elsif ($opt eq 'vcpus') {
5109 die "skip\n" if !$hotplug_features->{cpu
};
5110 qemu_cpu_hotplug
($vmid, $conf, $value);
5111 } elsif ($opt eq 'balloon') {
5112 # enable/disable balloning device is not hotpluggable
5113 my $old_balloon_enabled = !!(!defined($conf->{balloon
}) || $conf->{balloon
});
5114 my $new_balloon_enabled = !!(!defined($conf->{pending
}->{balloon
}) || $conf->{pending
}->{balloon
});
5115 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5117 # allow manual ballooning if shares is set to zero
5118 if ((defined($conf->{shares
}) && ($conf->{shares
} == 0))) {
5119 my $balloon = $conf->{pending
}->{balloon
} || $conf->{memory
} || $defaults->{memory
};
5120 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5122 } elsif ($opt =~ m/^net(\d+)$/) {
5123 # some changes can be done without hotplug
5124 vmconfig_update_net
($storecfg, $conf, $hotplug_features->{network
},
5125 $vmid, $opt, $value, $arch, $machine_type);
5126 } elsif (is_valid_drivename
($opt)) {
5127 die "skip\n" if $opt eq 'efidisk0' || $opt eq 'tpmstate0';
5128 # some changes can be done without hotplug
5129 my $drive = parse_drive
($opt, $value);
5130 if (drive_is_cloudinit
($drive)) {
5131 $cloudinit_opt = [$opt, $drive];
5132 # apply all the other changes first, then generate the cloudinit disk
5135 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5136 $vmid, $opt, $value, $arch, $machine_type);
5137 } elsif ($opt =~ m/^memory$/) { #dimms
5138 die "skip\n" if !$hotplug_features->{memory
};
5139 $value = PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults, $value);
5140 } elsif ($opt eq 'cpuunits') {
5141 my $new_cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{pending
}->{$opt}); #clamp
5142 $cgroup->change_cpu_shares($new_cpuunits);
5143 } elsif ($opt eq 'cpulimit') {
5144 my $cpulimit = $conf->{pending
}->{$opt} == 0 ?
-1 : int($conf->{pending
}->{$opt} * 100000);
5145 $cgroup->change_cpu_quota($cpulimit, 100000);
5146 } elsif ($opt eq 'agent') {
5147 vmconfig_update_agent
($conf, $opt, $value);
5149 die "skip\n"; # skip non-hot-pluggable options
5153 &$add_error($opt, $err) if $err ne "skip\n";
5155 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $value);
5156 $conf->{$opt} = $value;
5157 delete $conf->{pending
}->{$opt};
5161 if (defined($cloudinit_opt)) {
5162 my ($opt, $drive) = @$cloudinit_opt;
5163 my $value = $conf->{pending
}->{$opt};
5165 my $temp = {%$conf, $opt => $value};
5166 PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($temp, $vmid);
5167 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5168 $vmid, $opt, $value, $arch, $machine_type);
5171 &$add_error($opt, $err) if $err ne "skip\n";
5173 $conf->{$opt} = $value;
5174 delete $conf->{pending
}->{$opt};
5178 # unplug xhci controller if no usb device is left
5181 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
5182 next if !defined($conf->{"usb$i"});
5187 vm_deviceunplug
($vmid, $conf, 'xhci');
5191 PVE
::QemuConfig-
>write_config($vmid, $conf);
5193 if ($hotplug_features->{cloudinit
} && PVE
::QemuServer
::Cloudinit
::has_changes
($conf)) {
5194 PVE
::QemuServer
::vmconfig_update_cloudinit_drive
($storecfg, $conf, $vmid);
5198 sub try_deallocate_drive
{
5199 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5201 if (($force || $key =~ /^unused/) && !drive_is_cdrom
($drive, 1)) {
5202 my $volid = $drive->{file
};
5203 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
5204 my $sid = PVE
::Storage
::parse_volume_id
($volid);
5205 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5207 # check if the disk is really unused
5208 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5209 if PVE
::QemuServer
::Drive
::is_volume_in_use
($storecfg, $conf, $key, $volid);
5210 PVE
::Storage
::vdisk_free
($storecfg, $volid);
5213 # If vm is not owner of this disk remove from config
5221 sub vmconfig_delete_or_detach_drive
{
5222 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5224 my $drive = parse_drive
($opt, $conf->{$opt});
5226 my $rpcenv = PVE
::RPCEnvironment
::get
();
5227 my $authuser = $rpcenv->get_user();
5230 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5231 try_deallocate_drive
($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5233 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $drive);
5239 sub vmconfig_apply_pending
{
5240 my ($vmid, $conf, $storecfg, $errors, $skip_cloud_init) = @_;
5242 return if !scalar(keys %{$conf->{pending
}});
5244 my $add_apply_error = sub {
5245 my ($opt, $msg) = @_;
5246 my $err_msg = "unable to apply pending change $opt : $msg";
5247 $errors->{$opt} = $err_msg;
5253 my $pending_delete_hash = PVE
::QemuConfig-
>parse_pending_delete($conf->{pending
}->{delete});
5254 foreach my $opt (sort keys %$pending_delete_hash) {
5255 my $force = $pending_delete_hash->{$opt}->{force
};
5257 if ($opt =~ m/^unused/) {
5258 die "internal error";
5259 } elsif (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5260 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5264 $add_apply_error->($opt, $err);
5266 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5267 delete $conf->{$opt};
5271 PVE
::QemuConfig-
>cleanup_pending($conf);
5273 my $generate_cloudinit = $skip_cloud_init ?
0 : undef;
5275 foreach my $opt (keys %{$conf->{pending
}}) { # add/change
5276 next if $opt eq 'delete'; # just to be sure
5278 if (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5279 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, parse_drive
($opt, $conf->{$opt}))
5283 $add_apply_error->($opt, $err);
5286 if (is_valid_drivename
($opt)) {
5287 my $drive = parse_drive
($opt, $conf->{pending
}->{$opt});
5288 $generate_cloudinit //= 1 if drive_is_cloudinit
($drive);
5291 $conf->{$opt} = delete $conf->{pending
}->{$opt};
5295 # write all changes at once to avoid unnecessary i/o
5296 PVE
::QemuConfig-
>write_config($vmid, $conf);
5297 if ($generate_cloudinit) {
5298 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5299 # After successful generation and if there were changes to be applied, update the
5300 # config to drop the {cloudinit} entry.
5301 PVE
::QemuConfig-
>write_config($vmid, $conf);
5306 sub vmconfig_update_net
{
5307 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5309 my $newnet = parse_net
($value);
5311 if ($conf->{$opt}) {
5312 my $oldnet = parse_net
($conf->{$opt});
5314 if (safe_string_ne
($oldnet->{model
}, $newnet->{model
}) ||
5315 safe_string_ne
($oldnet->{macaddr
}, $newnet->{macaddr
}) ||
5316 safe_num_ne
($oldnet->{queues
}, $newnet->{queues
}) ||
5317 !($newnet->{bridge
} && $oldnet->{bridge
})) { # bridge/nat mode change
5319 # for non online change, we try to hot-unplug
5320 die "skip\n" if !$hotplug;
5321 vm_deviceunplug
($vmid, $conf, $opt);
5324 die "internal error" if $opt !~ m/net(\d+)/;
5325 my $iface = "tap${vmid}i$1";
5327 if (safe_string_ne
($oldnet->{bridge
}, $newnet->{bridge
}) ||
5328 safe_num_ne
($oldnet->{tag
}, $newnet->{tag
}) ||
5329 safe_string_ne
($oldnet->{trunks
}, $newnet->{trunks
}) ||
5330 safe_num_ne
($oldnet->{firewall
}, $newnet->{firewall
})) {
5331 PVE
::Network
::tap_unplug
($iface);
5334 PVE
::Network
::SDN
::Zones
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5336 PVE
::Network
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5338 } elsif (safe_num_ne
($oldnet->{rate
}, $newnet->{rate
})) {
5339 # Rate can be applied on its own but any change above needs to
5340 # include the rate in tap_plug since OVS resets everything.
5341 PVE
::Network
::tap_rate_limit
($iface, $newnet->{rate
});
5344 if (safe_string_ne
($oldnet->{link_down
}, $newnet->{link_down
})) {
5345 qemu_set_link_status
($vmid, $opt, !$newnet->{link_down
});
5353 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5359 sub vmconfig_update_agent
{
5360 my ($conf, $opt, $value) = @_;
5362 die "skip\n" if !$conf->{$opt};
5364 my $hotplug_options = { fstrim_cloned_disks
=> 1 };
5366 my $old_agent = parse_guest_agent
($conf);
5367 my $agent = parse_guest_agent
({$opt => $value});
5369 for my $option (keys %$agent) { # added/changed options
5370 next if defined($hotplug_options->{$option});
5371 die "skip\n" if safe_string_ne
($agent->{$option}, $old_agent->{$option});
5374 for my $option (keys %$old_agent) { # removed options
5375 next if defined($hotplug_options->{$option});
5376 die "skip\n" if safe_string_ne
($old_agent->{$option}, $agent->{$option});
5379 return; # either no actual change (e.g., format string reordered) or just hotpluggable changes
5382 sub vmconfig_update_disk
{
5383 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5385 my $drive = parse_drive
($opt, $value);
5387 if ($conf->{$opt} && (my $old_drive = parse_drive
($opt, $conf->{$opt}))) {
5388 my $media = $drive->{media
} || 'disk';
5389 my $oldmedia = $old_drive->{media
} || 'disk';
5390 die "unable to change media type\n" if $media ne $oldmedia;
5392 if (!drive_is_cdrom
($old_drive)) {
5394 if ($drive->{file
} ne $old_drive->{file
}) {
5396 die "skip\n" if !$hotplug;
5398 # unplug and register as unused
5399 vm_deviceunplug
($vmid, $conf, $opt);
5400 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive)
5403 # update existing disk
5405 # skip non hotpluggable value
5406 if (safe_string_ne
($drive->{aio
}, $old_drive->{aio
}) ||
5407 safe_string_ne
($drive->{discard
}, $old_drive->{discard
}) ||
5408 safe_string_ne
($drive->{iothread
}, $old_drive->{iothread
}) ||
5409 safe_string_ne
($drive->{queues
}, $old_drive->{queues
}) ||
5410 safe_string_ne
($drive->{cache
}, $old_drive->{cache
}) ||
5411 safe_string_ne
($drive->{ssd
}, $old_drive->{ssd
}) ||
5412 safe_string_ne
($drive->{ro
}, $old_drive->{ro
})) {
5417 if (safe_num_ne
($drive->{mbps
}, $old_drive->{mbps
}) ||
5418 safe_num_ne
($drive->{mbps_rd
}, $old_drive->{mbps_rd
}) ||
5419 safe_num_ne
($drive->{mbps_wr
}, $old_drive->{mbps_wr
}) ||
5420 safe_num_ne
($drive->{iops
}, $old_drive->{iops
}) ||
5421 safe_num_ne
($drive->{iops_rd
}, $old_drive->{iops_rd
}) ||
5422 safe_num_ne
($drive->{iops_wr
}, $old_drive->{iops_wr
}) ||
5423 safe_num_ne
($drive->{mbps_max
}, $old_drive->{mbps_max
}) ||
5424 safe_num_ne
($drive->{mbps_rd_max
}, $old_drive->{mbps_rd_max
}) ||
5425 safe_num_ne
($drive->{mbps_wr_max
}, $old_drive->{mbps_wr_max
}) ||
5426 safe_num_ne
($drive->{iops_max
}, $old_drive->{iops_max
}) ||
5427 safe_num_ne
($drive->{iops_rd_max
}, $old_drive->{iops_rd_max
}) ||
5428 safe_num_ne
($drive->{iops_wr_max
}, $old_drive->{iops_wr_max
}) ||
5429 safe_num_ne
($drive->{bps_max_length
}, $old_drive->{bps_max_length
}) ||
5430 safe_num_ne
($drive->{bps_rd_max_length
}, $old_drive->{bps_rd_max_length
}) ||
5431 safe_num_ne
($drive->{bps_wr_max_length
}, $old_drive->{bps_wr_max_length
}) ||
5432 safe_num_ne
($drive->{iops_max_length
}, $old_drive->{iops_max_length
}) ||
5433 safe_num_ne
($drive->{iops_rd_max_length
}, $old_drive->{iops_rd_max_length
}) ||
5434 safe_num_ne
($drive->{iops_wr_max_length
}, $old_drive->{iops_wr_max_length
})) {
5436 qemu_block_set_io_throttle
(
5438 ($drive->{mbps
} || 0)*1024*1024,
5439 ($drive->{mbps_rd
} || 0)*1024*1024,
5440 ($drive->{mbps_wr
} || 0)*1024*1024,
5441 $drive->{iops
} || 0,
5442 $drive->{iops_rd
} || 0,
5443 $drive->{iops_wr
} || 0,
5444 ($drive->{mbps_max
} || 0)*1024*1024,
5445 ($drive->{mbps_rd_max
} || 0)*1024*1024,
5446 ($drive->{mbps_wr_max
} || 0)*1024*1024,
5447 $drive->{iops_max
} || 0,
5448 $drive->{iops_rd_max
} || 0,
5449 $drive->{iops_wr_max
} || 0,
5450 $drive->{bps_max_length
} || 1,
5451 $drive->{bps_rd_max_length
} || 1,
5452 $drive->{bps_wr_max_length
} || 1,
5453 $drive->{iops_max_length
} || 1,
5454 $drive->{iops_rd_max_length
} || 1,
5455 $drive->{iops_wr_max_length
} || 1,
5465 if ($drive->{file
} eq 'none') {
5466 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5467 if (drive_is_cloudinit
($old_drive)) {
5468 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive);
5471 my $path = get_iso_path
($storecfg, $vmid, $drive->{file
});
5473 # force eject if locked
5474 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5477 mon_cmd
($vmid, "blockdev-change-medium",
5478 id
=> "$opt", filename
=> "$path");
5486 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5488 PVE
::Storage
::activate_volumes
($storecfg, [$drive->{file
}]) if $drive->{file
} !~ m
|^/dev/.+|;
5489 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5492 sub vmconfig_update_cloudinit_drive
{
5493 my ($storecfg, $conf, $vmid) = @_;
5495 my $cloudinit_ds = undef;
5496 my $cloudinit_drive = undef;
5498 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5499 my ($ds, $drive) = @_;
5500 if (PVE
::QemuServer
::drive_is_cloudinit
($drive)) {
5501 $cloudinit_ds = $ds;
5502 $cloudinit_drive = $drive;
5506 return if !$cloudinit_drive;
5508 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5509 PVE
::QemuConfig-
>write_config($vmid, $conf);
5512 my $running = PVE
::QemuServer
::check_running
($vmid);
5515 my $path = PVE
::Storage
::path
($storecfg, $cloudinit_drive->{file
});
5517 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$cloudinit_ds");
5518 mon_cmd
($vmid, "blockdev-change-medium", id
=> "$cloudinit_ds", filename
=> "$path");
5523 # called in locked context by incoming migration
5524 sub vm_migrate_get_nbd_disks
{
5525 my ($storecfg, $conf, $replicated_volumes) = @_;
5527 my $local_volumes = {};
5528 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5529 my ($ds, $drive) = @_;
5531 return if drive_is_cdrom
($drive);
5532 return if $ds eq 'tpmstate0';
5534 my $volid = $drive->{file
};
5538 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
5540 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5541 return if $scfg->{shared
};
5543 # replicated disks re-use existing state via bitmap
5544 my $use_existing = $replicated_volumes->{$volid} ?
1 : 0;
5545 $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
5547 return $local_volumes;
5550 # called in locked context by incoming migration
5551 sub vm_migrate_alloc_nbd_disks
{
5552 my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
5555 foreach my $opt (sort keys %$source_volumes) {
5556 my ($volid, $storeid, $volname, $drive, $use_existing, $format) = @{$source_volumes->{$opt}};
5558 if ($use_existing) {
5559 $nbd->{$opt}->{drivestr
} = print_drive
($drive);
5560 $nbd->{$opt}->{volid
} = $volid;
5561 $nbd->{$opt}->{replicated
} = 1;
5565 # storage mapping + volname = regular migration
5566 # storage mapping + format = remote migration
5567 # order of precedence, filtered by whether storage supports it:
5568 # 1. explicit requested format
5569 # 2. format of current volume
5570 # 3. default format of storage
5571 if (!$storagemap->{identity
}) {
5572 $storeid = PVE
::JSONSchema
::map_id
($storagemap, $storeid);
5573 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
5574 if (!$format || !grep { $format eq $_ } @$validFormats) {
5576 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5577 my $fileFormat = qemu_img_format
($scfg, $volname);
5578 $format = $fileFormat
5579 if grep { $fileFormat eq $_ } @$validFormats;
5581 $format //= $defFormat;
5584 # can't happen for remote migration, so $volname is always defined
5585 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5586 $format = qemu_img_format
($scfg, $volname);
5589 my $size = $drive->{size
} / 1024;
5590 my $newvolid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $format, undef, $size);
5591 my $newdrive = $drive;
5592 $newdrive->{format
} = $format;
5593 $newdrive->{file
} = $newvolid;
5594 my $drivestr = print_drive
($newdrive);
5595 $nbd->{$opt}->{drivestr
} = $drivestr;
5596 $nbd->{$opt}->{volid
} = $newvolid;
5602 # see vm_start_nolock for parameters, additionally:
5604 # storagemap = parsed storage map for allocating NBD disks
5606 my ($storecfg, $vmid, $params, $migrate_opts) = @_;
5608 return PVE
::QemuConfig-
>lock_config($vmid, sub {
5609 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migrate_opts->{migratedfrom
});
5611 die "you can't start a vm if it's a template\n"
5612 if !$params->{skiptemplate
} && PVE
::QemuConfig-
>is_template($conf);
5614 my $has_suspended_lock = PVE
::QemuConfig-
>has_lock($conf, 'suspended');
5615 my $has_backup_lock = PVE
::QemuConfig-
>has_lock($conf, 'backup');
5617 my $running = check_running
($vmid, undef, $migrate_opts->{migratedfrom
});
5619 if ($has_backup_lock && $running) {
5620 # a backup is currently running, attempt to start the guest in the
5621 # existing QEMU instance
5622 return vm_resume
($vmid);
5625 PVE
::QemuConfig-
>check_lock($conf)
5626 if !($params->{skiplock
} || $has_suspended_lock);
5628 $params->{resume
} = $has_suspended_lock || defined($conf->{vmstate
});
5630 die "VM $vmid already running\n" if $running;
5632 if (my $storagemap = $migrate_opts->{storagemap
}) {
5633 my $replicated = $migrate_opts->{replicated_volumes
};
5634 my $disks = vm_migrate_get_nbd_disks
($storecfg, $conf, $replicated);
5635 $migrate_opts->{nbd
} = vm_migrate_alloc_nbd_disks
($storecfg, $vmid, $disks, $storagemap);
5637 foreach my $opt (keys %{$migrate_opts->{nbd
}}) {
5638 $conf->{$opt} = $migrate_opts->{nbd
}->{$opt}->{drivestr
};
5642 return vm_start_nolock
($storecfg, $vmid, $conf, $params, $migrate_opts);
5648 # statefile => 'tcp', 'unix' for migration or path/volid for RAM state
5649 # skiplock => 0/1, skip checking for config lock
5650 # skiptemplate => 0/1, skip checking whether VM is template
5651 # forcemachine => to force QEMU machine (rollback/migration)
5652 # forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
5653 # timeout => in seconds
5654 # paused => start VM in paused state (backup)
5655 # resume => resume from hibernation
5666 # nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
5667 # migratedfrom => source node
5668 # spice_ticket => used for spice migration, passed via tunnel/stdin
5669 # network => CIDR of migration network
5670 # type => secure/insecure - tunnel over encrypted connection or plain-text
5671 # nbd_proto_version => int, 0 for TCP, 1 for UNIX
5672 # replicated_volumes => which volids should be re-used with bitmaps for nbd migration
5673 # offline_volumes => new volids of offline migrated disks like tpmstate and cloudinit, not yet
5674 # contained in config
5675 sub vm_start_nolock
{
5676 my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
5678 my $statefile = $params->{statefile
};
5679 my $resume = $params->{resume
};
5681 my $migratedfrom = $migrate_opts->{migratedfrom
};
5682 my $migration_type = $migrate_opts->{type
};
5686 # clean up leftover reboot request files
5687 eval { clear_reboot_request
($vmid); };
5690 if (!$statefile && scalar(keys %{$conf->{pending
}})) {
5691 vmconfig_apply_pending
($vmid, $conf, $storecfg);
5692 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5695 # don't regenerate the ISO if the VM is started as part of a live migration
5696 # this way we can reuse the old ISO with the correct config
5697 if (!$migratedfrom) {
5698 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5699 # FIXME: apply_cloudinit_config updates $conf in this case, and it would only drop
5700 # $conf->{cloudinit}, so we could just not do this?
5701 # But we do it above, so for now let's be consistent.
5702 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5706 # override offline migrated volumes, conf is out of date still
5707 if (my $offline_volumes = $migrate_opts->{offline_volumes
}) {
5708 for my $key (sort keys $offline_volumes->%*) {
5709 my $parsed = parse_drive
($key, $conf->{$key});
5710 $parsed->{file
} = $offline_volumes->{$key};
5711 $conf->{$key} = print_drive
($parsed);
5715 my $defaults = load_defaults
();
5717 # set environment variable useful inside network script
5718 # for remote migration the config is available on the target node!
5719 if (!$migrate_opts->{remote_node
}) {
5720 $ENV{PVE_MIGRATED_FROM
} = $migratedfrom;
5723 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-start', 1);
5725 my $forcemachine = $params->{forcemachine
};
5726 my $forcecpu = $params->{forcecpu
};
5728 # enforce machine and CPU type on suspended vm to ensure HW compatibility
5729 $forcemachine = $conf->{runningmachine
};
5730 $forcecpu = $conf->{runningcpu
};
5731 print "Resuming suspended VM\n";
5734 my ($cmd, $vollist, $spice_port) = config_to_command
($storecfg, $vmid,
5735 $conf, $defaults, $forcemachine, $forcecpu, $params->{'pbs-backing'});
5738 my $get_migration_ip = sub {
5739 my ($nodename) = @_;
5741 return $migration_ip if defined($migration_ip);
5743 my $cidr = $migrate_opts->{network
};
5745 if (!defined($cidr)) {
5746 my $dc_conf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5747 $cidr = $dc_conf->{migration
}->{network
};
5750 if (defined($cidr)) {
5751 my $ips = PVE
::Network
::get_local_ip_from_cidr
($cidr);
5753 die "could not get IP: no address configured on local " .
5754 "node for network '$cidr'\n" if scalar(@$ips) == 0;
5756 die "could not get IP: multiple addresses configured on local " .
5757 "node for network '$cidr'\n" if scalar(@$ips) > 1;
5759 $migration_ip = @$ips[0];
5762 $migration_ip = PVE
::Cluster
::remote_node_ip
($nodename, 1)
5763 if !defined($migration_ip);
5765 return $migration_ip;
5769 if ($statefile eq 'tcp') {
5770 my $migrate = $res->{migrate
} = { proto
=> 'tcp' };
5771 $migrate->{addr
} = "localhost";
5772 my $datacenterconf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5773 my $nodename = nodename
();
5775 if (!defined($migration_type)) {
5776 if (defined($datacenterconf->{migration
}->{type
})) {
5777 $migration_type = $datacenterconf->{migration
}->{type
};
5779 $migration_type = 'secure';
5783 if ($migration_type eq 'insecure') {
5784 $migrate->{addr
} = $get_migration_ip->($nodename);
5785 $migrate->{addr
} = "[$migrate->{addr}]" if Net
::IP
::ip_is_ipv6
($migrate->{addr
});
5788 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5789 $migrate->{port
} = PVE
::Tools
::next_migrate_port
($pfamily);
5790 $migrate->{uri
} = "tcp:$migrate->{addr}:$migrate->{port}";
5791 push @$cmd, '-incoming', $migrate->{uri
};
5794 } elsif ($statefile eq 'unix') {
5795 # should be default for secure migrations as a ssh TCP forward
5796 # tunnel is not deterministic reliable ready and fails regurarly
5797 # to set up in time, so use UNIX socket forwards
5798 my $migrate = $res->{migrate
} = { proto
=> 'unix' };
5799 $migrate->{addr
} = "/run/qemu-server/$vmid.migrate";
5800 unlink $migrate->{addr
};
5802 $migrate->{uri
} = "unix:$migrate->{addr}";
5803 push @$cmd, '-incoming', $migrate->{uri
};
5806 } elsif (-e
$statefile) {
5807 push @$cmd, '-loadstate', $statefile;
5809 my $statepath = PVE
::Storage
::path
($storecfg, $statefile);
5810 push @$vollist, $statefile;
5811 push @$cmd, '-loadstate', $statepath;
5813 } elsif ($params->{paused
}) {
5817 my $start_timeout = $params->{timeout
} // config_aware_timeout
($conf, $resume);
5819 my $pci_devices = {}; # host pci devices
5820 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
5821 my $dev = $conf->{"hostpci$i"} or next;
5822 $pci_devices->{$i} = parse_hostpci
($dev);
5825 # do not reserve pciid for mediated devices, sysfs will error out for duplicate assignment
5826 my $real_pci_devices = [ grep { !(defined($_->{mdev
}) && scalar($_->{pciid
}->@*) == 1) } values $pci_devices->%* ];
5828 # map to a flat list of pci ids
5829 my $pci_id_list = [ map { $_->{id
} } map { $_->{pciid
}->@* } $real_pci_devices->@* ];
5831 # reserve all PCI IDs before actually doing anything with them
5832 PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, $start_timeout);
5836 for my $id (sort keys %$pci_devices) {
5837 my $d = $pci_devices->{$id};
5838 for my $dev ($d->{pciid
}->@*) {
5839 my $info = PVE
::QemuServer
::PCI
::prepare_pci_device
($vmid, $dev->{id
}, $id, $d->{mdev
});
5841 # nvidia grid needs the uuid of the mdev as qemu parameter
5842 if ($d->{mdev
} && !defined($uuid) && $info->{vendor
} eq '10de') {
5843 $uuid = PVE
::QemuServer
::PCI
::generate_mdev_uuid
($vmid, $id);
5847 push @$cmd, '-uuid', $uuid if defined($uuid);
5850 eval { cleanup_pci_devices
($vmid, $conf) };
5855 PVE
::Storage
::activate_volumes
($storecfg, $vollist);
5858 run_command
(['/bin/systemctl', 'stop', "$vmid.scope"], outfunc
=> sub{}, errfunc
=> sub{});
5860 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5861 # timeout should be more than enough here...
5862 PVE
::Systemd
::wait_for_unit_removed
("$vmid.scope", 20);
5864 my $cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{cpuunits
});
5867 timeout
=> $statefile ?
undef : $start_timeout,
5872 # when migrating, prefix QEMU output so other side can pick up any
5873 # errors that might occur and show the user
5874 if ($migratedfrom) {
5875 $run_params{quiet
} = 1;
5876 $run_params{logfunc
} = sub { print "QEMU: $_[0]\n" };
5879 my %systemd_properties = (
5880 Slice
=> 'qemu.slice',
5881 KillMode
=> 'process',
5883 TimeoutStopUSec
=> ULONG_MAX
, # infinity
5886 if (PVE
::CGroup
::cgroup_mode
() == 2) {
5887 $systemd_properties{CPUWeight
} = $cpuunits;
5889 $systemd_properties{CPUShares
} = $cpuunits;
5892 if (my $cpulimit = $conf->{cpulimit
}) {
5893 $systemd_properties{CPUQuota
} = int($cpulimit * 100);
5895 $systemd_properties{timeout
} = 10 if $statefile; # setting up the scope shoul be quick
5897 my $run_qemu = sub {
5898 PVE
::Tools
::run_fork
sub {
5899 PVE
::Systemd
::enter_systemd_scope
($vmid, "Proxmox VE VM $vmid", %systemd_properties);
5902 if (my $tpm = $conf->{tpmstate0
}) {
5903 # start the TPM emulator so QEMU can connect on start
5904 $tpmpid = start_swtpm
($storecfg, $vmid, $tpm, $migratedfrom);
5907 my $exitcode = run_command
($cmd, %run_params);
5910 warn "stopping swtpm instance (pid $tpmpid) due to QEMU startup error\n";
5911 kill 'TERM', $tpmpid;
5913 die "QEMU exited with code $exitcode\n";
5918 if ($conf->{hugepages
}) {
5921 my $hugepages_topology = PVE
::QemuServer
::Memory
::hugepages_topology
($conf);
5922 my $hugepages_host_topology = PVE
::QemuServer
::Memory
::hugepages_host_topology
();
5924 PVE
::QemuServer
::Memory
::hugepages_mount
();
5925 PVE
::QemuServer
::Memory
::hugepages_allocate
($hugepages_topology, $hugepages_host_topology);
5927 eval { $run_qemu->() };
5929 PVE
::QemuServer
::Memory
::hugepages_reset
($hugepages_host_topology)
5930 if !$conf->{keephugepages
};
5934 PVE
::QemuServer
::Memory
::hugepages_pre_deallocate
($hugepages_topology)
5935 if !$conf->{keephugepages
};
5937 eval { PVE
::QemuServer
::Memory
::hugepages_update_locked
($code); };
5940 eval { $run_qemu->() };
5944 # deactivate volumes if start fails
5945 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
5947 eval { cleanup_pci_devices
($vmid, $conf) };
5950 die "start failed: $err";
5953 # re-reserve all PCI IDs now that we can know the actual VM PID
5954 my $pid = PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
5955 eval { PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, undef, $pid) };
5958 if (defined($res->{migrate
})) {
5959 print "migration listens on $res->{migrate}->{uri}\n";
5960 } elsif ($statefile) {
5961 eval { mon_cmd
($vmid, "cont"); };
5965 #start nbd server for storage migration
5966 if (my $nbd = $migrate_opts->{nbd
}) {
5967 my $nbd_protocol_version = $migrate_opts->{nbd_proto_version
} // 0;
5969 my $migrate_storage_uri;
5970 # nbd_protocol_version > 0 for unix socket support
5971 if ($nbd_protocol_version > 0 && ($migration_type eq 'secure' || $migration_type eq 'websocket')) {
5972 my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
5973 mon_cmd
($vmid, "nbd-server-start", addr
=> { type
=> 'unix', data
=> { path
=> $socket_path } } );
5974 $migrate_storage_uri = "nbd:unix:$socket_path";
5975 $res->{migrate
}->{unix_sockets
} = [$socket_path];
5977 my $nodename = nodename
();
5978 my $localip = $get_migration_ip->($nodename);
5979 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5980 my $storage_migrate_port = PVE
::Tools
::next_migrate_port
($pfamily);
5982 mon_cmd
($vmid, "nbd-server-start", addr
=> {
5985 host
=> "${localip}",
5986 port
=> "${storage_migrate_port}",
5989 $localip = "[$localip]" if Net
::IP
::ip_is_ipv6
($localip);
5990 $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
5993 my $block_info = mon_cmd
($vmid, "query-block");
5994 $block_info = { map { $_->{device
} => $_ } $block_info->@* };
5996 foreach my $opt (sort keys %$nbd) {
5997 my $drivestr = $nbd->{$opt}->{drivestr
};
5998 my $volid = $nbd->{$opt}->{volid
};
6000 my $block_node = $block_info->{"drive-$opt"}->{inserted
}->{'node-name'};
6006 'node-name' => $block_node,
6007 writable
=> JSON
::true
,
6009 name
=> "drive-$opt", # NBD export name
6012 my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
6013 print "storage migration listens on $nbd_uri volume:$drivestr\n";
6014 print "re-using replicated volume: $opt - $volid\n"
6015 if $nbd->{$opt}->{replicated
};
6017 $res->{drives
}->{$opt} = $nbd->{$opt};
6018 $res->{drives
}->{$opt}->{nbd_uri
} = $nbd_uri;
6022 if ($migratedfrom) {
6024 set_migration_caps
($vmid);
6029 print "spice listens on port $spice_port\n";
6030 $res->{spice_port
} = $spice_port;
6031 if ($migrate_opts->{spice_ticket
}) {
6032 mon_cmd
($vmid, "set_password", protocol
=> 'spice', password
=>
6033 $migrate_opts->{spice_ticket
});
6034 mon_cmd
($vmid, "expire_password", protocol
=> 'spice', time => "+30");
6039 mon_cmd
($vmid, "balloon", value
=> $conf->{balloon
}*1024*1024)
6040 if !$statefile && $conf->{balloon
};
6042 foreach my $opt (keys %$conf) {
6043 next if $opt !~ m/^net\d+$/;
6044 my $nicconf = parse_net
($conf->{$opt});
6045 qemu_set_link_status
($vmid, $opt, 0) if $nicconf->{link_down
};
6047 add_nets_bridge_fdb
($conf, $vmid);
6050 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
6054 path
=> "machine/peripheral/balloon0",
6055 property
=> "guest-stats-polling-interval",
6061 print "Resumed VM, removing state\n";
6062 if (my $vmstate = $conf->{vmstate
}) {
6063 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6064 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6066 delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
6067 PVE
::QemuConfig-
>write_config($vmid, $conf);
6070 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'post-start');
6075 sub vm_commandline
{
6076 my ($storecfg, $vmid, $snapname) = @_;
6078 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6080 my ($forcemachine, $forcecpu);
6082 my $snapshot = $conf->{snapshots
}->{$snapname};
6083 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
6085 # check for machine or CPU overrides in snapshot
6086 $forcemachine = $snapshot->{runningmachine
};
6087 $forcecpu = $snapshot->{runningcpu
};
6089 $snapshot->{digest
} = $conf->{digest
}; # keep file digest for API
6094 my $defaults = load_defaults
();
6096 my $cmd = config_to_command
($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
6098 return PVE
::Tools
::cmd2string
($cmd);
6102 my ($vmid, $skiplock) = @_;
6104 PVE
::QemuConfig-
>lock_config($vmid, sub {
6106 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6108 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6110 mon_cmd
($vmid, "system_reset");
6114 sub get_vm_volumes
{
6118 foreach_volid
($conf, sub {
6119 my ($volid, $attr) = @_;
6121 return if $volid =~ m
|^/|;
6123 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
6126 push @$vollist, $volid;
6132 sub cleanup_pci_devices
{
6133 my ($vmid, $conf) = @_;
6135 foreach my $key (keys %$conf) {
6136 next if $key !~ m/^hostpci(\d+)$/;
6137 my $hostpciindex = $1;
6138 my $uuid = PVE
::SysFSTools
::generate_mdev_uuid
($vmid, $hostpciindex);
6139 my $d = parse_hostpci
($conf->{$key});
6141 # NOTE: avoid PVE::SysFSTools::pci_cleanup_mdev_device as it requires PCI ID and we
6142 # don't want to break ABI just for this two liner
6143 my $dev_sysfs_dir = "/sys/bus/mdev/devices/$uuid";
6144 PVE
::SysFSTools
::file_write
("$dev_sysfs_dir/remove", "1") if -e
$dev_sysfs_dir;
6147 PVE
::QemuServer
::PCI
::remove_pci_reservation
($vmid);
6150 sub vm_stop_cleanup
{
6151 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
6156 my $vollist = get_vm_volumes
($conf);
6157 PVE
::Storage
::deactivate_volumes
($storecfg, $vollist);
6159 if (my $tpmdrive = $conf->{tpmstate0
}) {
6160 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
6161 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
6163 PVE
::Storage
::unmap_volume
($storecfg, $tpm->{file
});
6168 foreach my $ext (qw(mon qmp pid vnc qga)) {
6169 unlink "/var/run/qemu-server/${vmid}.$ext";
6172 if ($conf->{ivshmem
}) {
6173 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
6174 # just delete it for now, VMs which have this already open do not
6175 # are affected, but new VMs will get a separated one. If this
6176 # becomes an issue we either add some sort of ref-counting or just
6177 # add a "don't delete on stop" flag to the ivshmem format.
6178 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name
} // $vmid);
6181 cleanup_pci_devices
($vmid, $conf);
6183 vmconfig_apply_pending
($vmid, $conf, $storecfg) if $apply_pending_changes;
6185 warn $@ if $@; # avoid errors - just warn
6188 # call only in locked context
6190 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
6192 my $pid = check_running
($vmid, $nocheck);
6197 $conf = PVE
::QemuConfig-
>load_config($vmid);
6198 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6199 if (!defined($timeout) && $shutdown && $conf->{startup
}) {
6200 my $opts = PVE
::JSONSchema
::pve_parse_startup_order
($conf->{startup
});
6201 $timeout = $opts->{down
} if $opts->{down
};
6203 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-stop');
6208 if (defined($conf) && get_qga_key
($conf, 'enabled')) {
6209 mon_cmd
($vmid, "guest-shutdown", timeout
=> $timeout);
6211 mon_cmd
($vmid, "system_powerdown");
6214 mon_cmd
($vmid, "quit");
6220 $timeout = 60 if !defined($timeout);
6223 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6228 if ($count >= $timeout) {
6230 warn "VM still running - terminating now with SIGTERM\n";
6233 die "VM quit/powerdown failed - got timeout\n";
6236 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6240 if (!check_running
($vmid, $nocheck)) {
6241 warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
6245 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
6248 die "VM quit/powerdown failed\n";
6256 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6261 if ($count >= $timeout) {
6262 warn "VM still running - terminating now with SIGKILL\n";
6267 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6270 # Note: use $nocheck to skip tests if VM configuration file exists.
6271 # We need that when migration VMs to other nodes (files already moved)
6272 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
6274 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
6276 $force = 1 if !defined($force) && !$shutdown;
6279 my $pid = check_running
($vmid, $nocheck, $migratedfrom);
6280 kill 15, $pid if $pid;
6281 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migratedfrom);
6282 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 0);
6286 PVE
::QemuConfig-
>lock_config($vmid, sub {
6287 _do_vm_stop
($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
6292 my ($vmid, $timeout) = @_;
6294 PVE
::QemuConfig-
>lock_config($vmid, sub {
6297 # only reboot if running, as qmeventd starts it again on a stop event
6298 return if !check_running
($vmid);
6300 create_reboot_request
($vmid);
6302 my $storecfg = PVE
::Storage
::config
();
6303 _do_vm_stop
($storecfg, $vmid, undef, undef, $timeout, 1);
6307 # avoid that the next normal shutdown will be confused for a reboot
6308 clear_reboot_request
($vmid);
6314 # note: if using the statestorage parameter, the caller has to check privileges
6316 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
6323 PVE
::QemuConfig-
>lock_config($vmid, sub {
6325 $conf = PVE
::QemuConfig-
>load_config($vmid);
6327 my $is_backing_up = PVE
::QemuConfig-
>has_lock($conf, 'backup');
6328 PVE
::QemuConfig-
>check_lock($conf)
6329 if !($skiplock || $is_backing_up);
6331 die "cannot suspend to disk during backup\n"
6332 if $is_backing_up && $includestate;
6334 if ($includestate) {
6335 $conf->{lock} = 'suspending';
6336 my $date = strftime
("%Y-%m-%d", localtime(time()));
6337 $storecfg = PVE
::Storage
::config
();
6338 if (!$statestorage) {
6339 $statestorage = find_vmstate_storage
($conf, $storecfg);
6340 # check permissions for the storage
6341 my $rpcenv = PVE
::RPCEnvironment
::get
();
6342 if ($rpcenv->{type
} ne 'cli') {
6343 my $authuser = $rpcenv->get_user();
6344 $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
6349 $vmstate = PVE
::QemuConfig-
>__snapshot_save_vmstate(
6350 $vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
6351 $path = PVE
::Storage
::path
($storecfg, $vmstate);
6352 PVE
::QemuConfig-
>write_config($vmid, $conf);
6354 mon_cmd
($vmid, "stop");
6358 if ($includestate) {
6360 PVE
::Storage
::activate_volumes
($storecfg, [$vmstate]);
6363 set_migration_caps
($vmid, 1);
6364 mon_cmd
($vmid, "savevm-start", statefile
=> $path);
6366 my $state = mon_cmd
($vmid, "query-savevm");
6367 if (!$state->{status
}) {
6368 die "savevm not active\n";
6369 } elsif ($state->{status
} eq 'active') {
6372 } elsif ($state->{status
} eq 'completed') {
6373 print "State saved, quitting\n";
6375 } elsif ($state->{status
} eq 'failed' && $state->{error
}) {
6376 die "query-savevm failed with error '$state->{error}'\n"
6378 die "query-savevm returned status '$state->{status}'\n";
6384 PVE
::QemuConfig-
>lock_config($vmid, sub {
6385 $conf = PVE
::QemuConfig-
>load_config($vmid);
6387 # cleanup, but leave suspending lock, to indicate something went wrong
6389 mon_cmd
($vmid, "savevm-end");
6390 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6391 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6392 delete $conf->@{qw(vmstate runningmachine runningcpu)};
6393 PVE
::QemuConfig-
>write_config($vmid, $conf);
6399 die "lock changed unexpectedly\n"
6400 if !PVE
::QemuConfig-
>has_lock($conf, 'suspending');
6402 mon_cmd
($vmid, "quit");
6403 $conf->{lock} = 'suspended';
6404 PVE
::QemuConfig-
>write_config($vmid, $conf);
6409 # $nocheck is set when called as part of a migration - in this context the
6410 # location of the config file (source or target node) is not deterministic,
6411 # since migration cannot wait for pmxcfs to process the rename
6413 my ($vmid, $skiplock, $nocheck) = @_;
6415 PVE
::QemuConfig-
>lock_config($vmid, sub {
6416 my $res = mon_cmd
($vmid, 'query-status');
6417 my $resume_cmd = 'cont';
6421 $conf = eval { PVE
::QemuConfig-
>load_config($vmid) }; # try on target node
6423 my $vmlist = PVE
::Cluster
::get_vmlist
();
6424 if (exists($vmlist->{ids
}->{$vmid})) {
6425 my $node = $vmlist->{ids
}->{$vmid}->{node
};
6426 $conf = eval { PVE
::QemuConfig-
>load_config($vmid, $node) }; # try on source node
6429 PVE
::Cluster
::cfs_update
(); # vmlist was wrong, invalidate cache
6430 $conf = PVE
::QemuConfig-
>load_config($vmid); # last try on target node again
6434 $conf = PVE
::QemuConfig-
>load_config($vmid);
6437 if ($res->{status
}) {
6438 return if $res->{status
} eq 'running'; # job done, go home
6439 $resume_cmd = 'system_wakeup' if $res->{status
} eq 'suspended';
6440 $reset = 1 if $res->{status
} eq 'shutdown';
6444 PVE
::QemuConfig-
>check_lock($conf)
6445 if !($skiplock || PVE
::QemuConfig-
>has_lock($conf, 'backup'));
6449 # required if a VM shuts down during a backup and we get a resume
6450 # request before the backup finishes for example
6451 mon_cmd
($vmid, "system_reset");
6454 add_nets_bridge_fdb
($conf, $vmid) if $resume_cmd eq 'cont';
6456 mon_cmd
($vmid, $resume_cmd);
6461 my ($vmid, $skiplock, $key) = @_;
6463 PVE
::QemuConfig-
>lock_config($vmid, sub {
6465 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6467 # there is no qmp command, so we use the human monitor command
6468 my $res = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "sendkey $key");
6469 die $res if $res ne '';
6473 # vzdump restore implementaion
6475 sub tar_archive_read_firstfile
{
6476 my $archive = shift;
6478 die "ERROR: file '$archive' does not exist\n" if ! -f
$archive;
6480 # try to detect archive type first
6481 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6482 die "unable to open file '$archive'\n";
6483 my $firstfile = <$fh>;
6487 die "ERROR: archive contaions no data\n" if !$firstfile;
6493 sub tar_restore_cleanup
{
6494 my ($storecfg, $statfile) = @_;
6496 print STDERR
"starting cleanup\n";
6498 if (my $fd = IO
::File-
>new($statfile, "r")) {
6499 while (defined(my $line = <$fd>)) {
6500 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6503 if ($volid =~ m
|^/|) {
6504 unlink $volid || die 'unlink failed\n';
6506 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6508 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6510 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6512 print STDERR
"unable to parse line in statfile - $line";
6519 sub restore_file_archive
{
6520 my ($archive, $vmid, $user, $opts) = @_;
6522 return restore_vma_archive
($archive, $vmid, $user, $opts)
6525 my $info = PVE
::Storage
::archive_info
($archive);
6526 my $format = $opts->{format
} // $info->{format
};
6527 my $comp = $info->{compression
};
6529 # try to detect archive format
6530 if ($format eq 'tar') {
6531 return restore_tar_archive
($archive, $vmid, $user, $opts);
6533 return restore_vma_archive
($archive, $vmid, $user, $opts, $comp);
6537 # hepler to remove disks that will not be used after restore
6538 my $restore_cleanup_oldconf = sub {
6539 my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
6541 my $kept_disks = {};
6543 PVE
::QemuConfig-
>foreach_volume($oldconf, sub {
6544 my ($ds, $drive) = @_;
6546 return if drive_is_cdrom
($drive, 1);
6548 my $volid = $drive->{file
};
6549 return if !$volid || $volid =~ m
|^/|;
6551 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
6552 return if !$path || !$owner || ($owner != $vmid);
6554 # Note: only delete disk we want to restore
6555 # other volumes will become unused
6556 if ($virtdev_hash->{$ds}) {
6557 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
6562 $kept_disks->{$volid} = 1;
6566 # after the restore we have no snapshots anymore
6567 for my $snapname (keys $oldconf->{snapshots
}->%*) {
6568 my $snap = $oldconf->{snapshots
}->{$snapname};
6569 if ($snap->{vmstate
}) {
6570 eval { PVE
::Storage
::vdisk_free
($storecfg, $snap->{vmstate
}); };
6576 for my $volid (keys $kept_disks->%*) {
6577 eval { PVE
::Storage
::volume_snapshot_delete
($storecfg, $volid, $snapname); };
6583 # Helper to parse vzdump backup device hints
6585 # $rpcenv: Environment, used to ckeck storage permissions
6586 # $user: User ID, to check storage permissions
6587 # $storecfg: Storage configuration
6588 # $fh: the file handle for reading the configuration
6589 # $devinfo: should contain device sizes for all backu-up'ed devices
6590 # $options: backup options (pool, default storage)
6592 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
6593 my $parse_backup_hints = sub {
6594 my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
6596 my $check_storage = sub { # assert if an image can be allocate
6597 my ($storeid, $scfg) = @_;
6598 die "Content type 'images' is not available on storage '$storeid'\n"
6599 if !$scfg->{content
}->{images
};
6600 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace'])
6601 if $user ne 'root@pam';
6604 my $virtdev_hash = {};
6605 while (defined(my $line = <$fh>)) {
6606 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6607 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6608 die "archive does not contain data for drive '$virtdev'\n"
6609 if !$devinfo->{$devname};
6611 if (defined($options->{storage
})) {
6612 $storeid = $options->{storage
} || 'local';
6613 } elsif (!$storeid) {
6616 $format = 'raw' if !$format;
6617 $devinfo->{$devname}->{devname
} = $devname;
6618 $devinfo->{$devname}->{virtdev
} = $virtdev;
6619 $devinfo->{$devname}->{format
} = $format;
6620 $devinfo->{$devname}->{storeid
} = $storeid;
6622 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6623 $check_storage->($storeid, $scfg); # permission and content type check
6625 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6626 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6628 my $drive = parse_drive
($virtdev, $2);
6630 if (drive_is_cloudinit
($drive)) {
6631 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
6632 $storeid = $options->{storage
} if defined ($options->{storage
});
6633 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6634 my $format = qemu_img_format
($scfg, $volname); # has 'raw' fallback
6636 $check_storage->($storeid, $scfg); # permission and content type check
6638 $virtdev_hash->{$virtdev} = {
6640 storeid
=> $storeid,
6641 size
=> PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
,
6648 return $virtdev_hash;
6651 # Helper to allocate and activate all volumes required for a restore
6653 # $storecfg: Storage configuration
6654 # $virtdev_hash: as returned by parse_backup_hints()
6656 # Returns: { $virtdev => $volid }
6657 my $restore_allocate_devices = sub {
6658 my ($storecfg, $virtdev_hash, $vmid) = @_;
6661 foreach my $virtdev (sort keys %$virtdev_hash) {
6662 my $d = $virtdev_hash->{$virtdev};
6663 my $alloc_size = int(($d->{size
} + 1024 - 1)/1024);
6664 my $storeid = $d->{storeid
};
6665 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6667 # test if requested format is supported
6668 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
6669 my $supported = grep { $_ eq $d->{format
} } @$validFormats;
6670 $d->{format
} = $defFormat if !$supported;
6673 if ($d->{is_cloudinit
}) {
6674 $name = "vm-$vmid-cloudinit";
6675 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6676 if ($scfg->{path
}) {
6677 $name .= ".$d->{format}";
6681 my $volid = PVE
::Storage
::vdisk_alloc
(
6682 $storecfg, $storeid, $vmid, $d->{format
}, $name, $alloc_size);
6684 print STDERR
"new volume ID is '$volid'\n";
6685 $d->{volid
} = $volid;
6687 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
6689 $map->{$virtdev} = $volid;
6695 sub restore_update_config_line
{
6696 my ($cookie, $map, $line, $unique) = @_;
6698 return '' if $line =~ m/^\#qmdump\#/;
6699 return '' if $line =~ m/^\#vzdump\#/;
6700 return '' if $line =~ m/^lock:/;
6701 return '' if $line =~ m/^unused\d+:/;
6702 return '' if $line =~ m/^parent:/;
6706 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
6707 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6708 # try to convert old 1.X settings
6709 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6710 foreach my $devconfig (PVE
::Tools
::split_list
($ethcfg)) {
6711 my ($model, $macaddr) = split(/\=/, $devconfig);
6712 $macaddr = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if !$macaddr || $unique;
6715 bridge
=> "vmbr$ind",
6716 macaddr
=> $macaddr,
6718 my $netstr = print_net
($net);
6720 $res .= "net$cookie->{netcount}: $netstr\n";
6721 $cookie->{netcount
}++;
6723 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6724 my ($id, $netstr) = ($1, $2);
6725 my $net = parse_net
($netstr);
6726 $net->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if $net->{macaddr
};
6727 $netstr = print_net
($net);
6728 $res .= "$id: $netstr\n";
6729 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk|tpmstate)\d+):\s*(\S+)\s*$/) {
6732 my $di = parse_drive
($virtdev, $value);
6733 if (defined($di->{backup
}) && !$di->{backup
}) {
6735 } elsif ($map->{$virtdev}) {
6736 delete $di->{format
}; # format can change on restore
6737 $di->{file
} = $map->{$virtdev};
6738 $value = print_drive
($di);
6739 $res .= "$virtdev: $value\n";
6743 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6745 if ($vmgenid ne '0') {
6746 # always generate a new vmgenid if there was a valid one setup
6747 $vmgenid = generate_uuid
();
6749 $res .= "vmgenid: $vmgenid\n";
6750 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6751 my ($uuid, $uuid_str);
6752 UUID
::generate
($uuid);
6753 UUID
::unparse
($uuid, $uuid_str);
6754 my $smbios1 = parse_smbios1
($2);
6755 $smbios1->{uuid
} = $uuid_str;
6756 $res .= $1.print_smbios1
($smbios1)."\n";
6764 my $restore_deactivate_volumes = sub {
6765 my ($storecfg, $virtdev_hash) = @_;
6768 for my $dev (values $virtdev_hash->%*) {
6769 push $vollist->@*, $dev->{volid
} if $dev->{volid
};
6772 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
6773 print STDERR
$@ if $@;
6776 my $restore_destroy_volumes = sub {
6777 my ($storecfg, $virtdev_hash) = @_;
6779 for my $dev (values $virtdev_hash->%*) {
6780 my $volid = $dev->{volid
} or next;
6782 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6783 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6785 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6789 my $restore_merge_config = sub {
6790 my ($filename, $backup_conf_raw, $override_conf) = @_;
6792 my $backup_conf = parse_vm_config
($filename, $backup_conf_raw);
6793 for my $key (keys $override_conf->%*) {
6794 $backup_conf->{$key} = $override_conf->{$key};
6797 return $backup_conf;
6801 my ($cfg, $vmid) = @_;
6803 my $info = PVE
::Storage
::vdisk_list
($cfg, undef, $vmid, undef, 'images');
6805 my $volid_hash = {};
6806 foreach my $storeid (keys %$info) {
6807 foreach my $item (@{$info->{$storeid}}) {
6808 next if !($item->{volid
} && $item->{size
});
6809 $item->{path
} = PVE
::Storage
::path
($cfg, $item->{volid
});
6810 $volid_hash->{$item->{volid
}} = $item;
6817 sub update_disk_config
{
6818 my ($vmid, $conf, $volid_hash) = @_;
6821 my $prefix = "VM $vmid";
6823 # used and unused disks
6824 my $referenced = {};
6826 # Note: it is allowed to define multiple storages with same path (alias), so
6827 # we need to check both 'volid' and real 'path' (two different volid can point
6828 # to the same path).
6830 my $referencedpath = {};
6833 PVE
::QemuConfig-
>foreach_volume($conf, sub {
6834 my ($opt, $drive) = @_;
6836 my $volid = $drive->{file
};
6838 my $volume = $volid_hash->{$volid};
6840 # mark volid as "in-use" for next step
6841 $referenced->{$volid} = 1;
6842 if ($volume && (my $path = $volume->{path
})) {
6843 $referencedpath->{$path} = 1;
6846 return if drive_is_cdrom
($drive);
6849 my ($updated, $msg) = PVE
::QemuServer
::Drive
::update_disksize
($drive, $volume->{size
});
6850 if (defined($updated)) {
6852 $conf->{$opt} = print_drive
($updated);
6853 print "$prefix ($opt): $msg\n";
6857 # remove 'unusedX' entry if volume is used
6858 PVE
::QemuConfig-
>foreach_unused_volume($conf, sub {
6859 my ($opt, $drive) = @_;
6861 my $volid = $drive->{file
};
6865 $path = $volid_hash->{$volid}->{path
} if $volid_hash->{$volid};
6866 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6867 print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
6869 delete $conf->{$opt};
6872 $referenced->{$volid} = 1;
6873 $referencedpath->{$path} = 1 if $path;
6876 foreach my $volid (sort keys %$volid_hash) {
6877 next if $volid =~ m/vm-$vmid-state-/;
6878 next if $referenced->{$volid};
6879 my $path = $volid_hash->{$volid}->{path
};
6880 next if !$path; # just to be sure
6881 next if $referencedpath->{$path};
6883 my $key = PVE
::QemuConfig-
>add_unused_volume($conf, $volid);
6884 print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
6885 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6892 my ($vmid, $nolock, $dryrun) = @_;
6894 my $cfg = PVE
::Storage
::config
();
6896 print "rescan volumes...\n";
6897 my $volid_hash = scan_volids
($cfg, $vmid);
6899 my $updatefn = sub {
6902 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6904 PVE
::QemuConfig-
>check_lock($conf);
6907 foreach my $volid (keys %$volid_hash) {
6908 my $info = $volid_hash->{$volid};
6909 $vm_volids->{$volid} = $info if $info->{vmid
} && $info->{vmid
} == $vmid;
6912 my $changes = update_disk_config
($vmid, $conf, $vm_volids);
6914 PVE
::QemuConfig-
>write_config($vmid, $conf) if $changes && !$dryrun;
6917 if (defined($vmid)) {
6921 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6924 my $vmlist = config_list
();
6925 foreach my $vmid (keys %$vmlist) {
6929 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6935 sub restore_proxmox_backup_archive
{
6936 my ($archive, $vmid, $user, $options) = @_;
6938 my $storecfg = PVE
::Storage
::config
();
6940 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($archive);
6941 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6943 my $fingerprint = $scfg->{fingerprint
};
6944 my $keyfile = PVE
::Storage
::PBSPlugin
::pbs_encryption_key_file_name
($storecfg, $storeid);
6946 my $repo = PVE
::PBSClient
::get_repository
($scfg);
6947 my $namespace = $scfg->{namespace
};
6949 # This is only used for `pbs-restore` and the QEMU PBS driver (live-restore)
6950 my $password = PVE
::Storage
::PBSPlugin
::pbs_get_password
($scfg, $storeid);
6951 local $ENV{PBS_PASSWORD
} = $password;
6952 local $ENV{PBS_FINGERPRINT
} = $fingerprint if defined($fingerprint);
6954 my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6955 PVE
::Storage
::parse_volname
($storecfg, $archive);
6957 die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
6959 die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
6961 my $tmpdir = "/var/tmp/vzdumptmp$$";
6965 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
6966 # disable interrupts (always do cleanups)
6970 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
6972 # Note: $oldconf is undef if VM does not exists
6973 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
6974 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
6975 my $new_conf_raw = '';
6977 my $rpcenv = PVE
::RPCEnvironment
::get
();
6978 my $devinfo = {}; # info about drives included in backup
6979 my $virtdev_hash = {}; # info about allocated drives
6987 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
6989 my $cfgfn = "$tmpdir/qemu-server.conf";
6990 my $firewall_config_fn = "$tmpdir/fw.conf";
6991 my $index_fn = "$tmpdir/index.json";
6993 my $cmd = "restore";
6995 my $param = [$pbs_backup_name, "index.json", $index_fn];
6996 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
6997 my $index = PVE
::Tools
::file_get_contents
($index_fn);
6998 $index = decode_json
($index);
7000 foreach my $info (@{$index->{files
}}) {
7001 if ($info->{filename
} =~ m/^(drive-\S+).img.fidx$/) {
7003 if ($info->{size
} =~ m/^(\d+)$/) { # untaint size
7004 $devinfo->{$devname}->{size
} = $1;
7006 die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
7011 my $is_qemu_server_backup = scalar(
7012 grep { $_->{filename
} eq 'qemu-server.conf.blob' } @{$index->{files
}}
7014 if (!$is_qemu_server_backup) {
7015 die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
7017 my $has_firewall_config = scalar(grep { $_->{filename
} eq 'fw.conf.blob' } @{$index->{files
}});
7019 $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
7020 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7022 if ($has_firewall_config) {
7023 $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
7024 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7026 my $pve_firewall_dir = '/etc/pve/firewall';
7027 mkdir $pve_firewall_dir; # make sure the dir exists
7028 PVE
::Tools
::file_copy
($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
7031 my $fh = IO
::File-
>new($cfgfn, "r") ||
7032 die "unable to read qemu-server.conf - $!\n";
7034 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
7036 # fixme: rate limit?
7038 # create empty/temp config
7039 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\nlock: create");
7041 $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
7044 my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
7046 foreach my $virtdev (sort keys %$virtdev_hash) {
7047 my $d = $virtdev_hash->{$virtdev};
7048 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7050 # this fails if storage is unavailable
7051 my $volid = $d->{volid
};
7052 my $path = PVE
::Storage
::path
($storecfg, $volid);
7054 # for live-restore we only want to preload the efidisk and TPM state
7055 next if $options->{live
} && $virtdev ne 'efidisk0' && $virtdev ne 'tpmstate0';
7058 if (defined(my $ns = $scfg->{namespace
})) {
7059 @ns_arg = ('--ns', $ns);
7062 my $pbs_restore_cmd = [
7063 '/usr/bin/pbs-restore',
7064 '--repository', $repo,
7067 "$d->{devname}.img.fidx",
7072 push @$pbs_restore_cmd, '--format', $d->{format
} if $d->{format
};
7073 push @$pbs_restore_cmd, '--keyfile', $keyfile if -e
$keyfile;
7075 if (PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $volid)) {
7076 push @$pbs_restore_cmd, '--skip-zero';
7079 my $dbg_cmdstring = PVE
::Tools
::cmd2string
($pbs_restore_cmd);
7080 print "restore proxmox backup image: $dbg_cmdstring\n";
7081 run_command
($pbs_restore_cmd);
7084 $fh->seek(0, 0) || die "seek failed - $!\n";
7086 my $cookie = { netcount
=> 0 };
7087 while (defined(my $line = <$fh>)) {
7088 $new_conf_raw .= restore_update_config_line
(
7100 if ($err || !$options->{live
}) {
7101 $restore_deactivate_volumes->($storecfg, $virtdev_hash);
7107 $restore_destroy_volumes->($storecfg, $virtdev_hash);
7111 if ($options->{live
}) {
7112 # keep lock during live-restore
7113 $new_conf_raw .= "\nlock: create";
7116 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $options->{override_conf
});
7117 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7119 eval { rescan
($vmid, 1); };
7122 PVE
::AccessControl
::add_vm_to_pool
($vmid, $options->{pool
}) if $options->{pool
};
7124 if ($options->{live
}) {
7130 local $SIG{PIPE
} = sub { die "got signal ($!) - abort\n"; };
7132 my $conf = PVE
::QemuConfig-
>load_config($vmid);
7133 die "cannot do live-restore for template\n" if PVE
::QemuConfig-
>is_template($conf);
7135 # these special drives are already restored before start
7136 delete $devinfo->{'drive-efidisk0'};
7137 delete $devinfo->{'drive-tpmstate0-backup'};
7141 keyfile
=> $keyfile,
7142 snapshot
=> $pbs_backup_name,
7143 namespace
=> $namespace,
7145 pbs_live_restore
($vmid, $conf, $storecfg, $devinfo, $pbs_opts);
7147 PVE
::QemuConfig-
>remove_lock($vmid, "create");
7151 sub pbs_live_restore
{
7152 my ($vmid, $conf, $storecfg, $restored_disks, $opts) = @_;
7154 print "starting VM for live-restore\n";
7155 print "repository: '$opts->{repo}', snapshot: '$opts->{snapshot}'\n";
7157 my $pbs_backing = {};
7158 for my $ds (keys %$restored_disks) {
7159 $ds =~ m/^drive-(.*)$/;
7161 $pbs_backing->{$confname} = {
7162 repository
=> $opts->{repo
},
7163 snapshot
=> $opts->{snapshot
},
7164 archive
=> "$ds.img.fidx",
7166 $pbs_backing->{$confname}->{keyfile
} = $opts->{keyfile
} if -e
$opts->{keyfile
};
7167 $pbs_backing->{$confname}->{namespace
} = $opts->{namespace
} if defined($opts->{namespace
});
7169 my $drive = parse_drive
($confname, $conf->{$confname});
7170 print "restoring '$ds' to '$drive->{file}'\n";
7173 my $drives_streamed = 0;
7175 # make sure HA doesn't interrupt our restore by stopping the VM
7176 if (PVE
::HA
::Config
::vm_is_ha_managed
($vmid)) {
7177 run_command
(['ha-manager', 'set', "vm:$vmid", '--state', 'started']);
7180 # start VM with backing chain pointing to PBS backup, environment vars for PBS driver
7181 # in QEMU (PBS_PASSWORD and PBS_FINGERPRINT) are already set by our caller
7182 vm_start_nolock
($storecfg, $vmid, $conf, {paused
=> 1, 'pbs-backing' => $pbs_backing}, {});
7184 my $qmeventd_fd = register_qmeventd_handle
($vmid);
7186 # begin streaming, i.e. data copy from PBS to target disk for every vol,
7187 # this will effectively collapse the backing image chain consisting of
7188 # [target <- alloc-track -> PBS snapshot] to just [target] (alloc-track
7189 # removes itself once all backing images vanish with 'auto-remove=on')
7191 for my $ds (sort keys %$restored_disks) {
7192 my $job_id = "restore-$ds";
7193 mon_cmd
($vmid, 'block-stream',
7194 'job-id' => $job_id,
7197 $jobs->{$job_id} = {};
7200 mon_cmd
($vmid, 'cont');
7201 qemu_drive_mirror_monitor
($vmid, undef, $jobs, 'auto', 0, 'stream');
7203 print "restore-drive jobs finished successfully, removing all tracking block devices"
7204 ." to disconnect from Proxmox Backup Server\n";
7206 for my $ds (sort keys %$restored_disks) {
7207 mon_cmd
($vmid, 'blockdev-del', 'node-name' => "$ds-pbs");
7210 close($qmeventd_fd);
7216 warn "An error occurred during live-restore: $err\n";
7217 _do_vm_stop
($storecfg, $vmid, 1, 1, 10, 0, 1);
7218 die "live-restore failed\n";
7222 sub restore_vma_archive
{
7223 my ($archive, $vmid, $user, $opts, $comp) = @_;
7225 my $readfrom = $archive;
7227 my $cfg = PVE
::Storage
::config
();
7229 my $bwlimit = $opts->{bwlimit
};
7231 my $dbg_cmdstring = '';
7232 my $add_pipe = sub {
7234 push @$commands, $cmd;
7235 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
7236 $dbg_cmdstring .= PVE
::Tools
::cmd2string
($cmd);
7241 if ($archive eq '-') {
7244 # If we use a backup from a PVE defined storage we also consider that
7245 # storage's rate limit:
7246 my (undef, $volid) = PVE
::Storage
::path_to_volume_id
($cfg, $archive);
7247 if (defined($volid)) {
7248 my ($sid, undef) = PVE
::Storage
::parse_volume_id
($volid);
7249 my $readlimit = PVE
::Storage
::get_bandwidth_limit
('restore', [$sid], $bwlimit);
7251 print STDERR
"applying read rate limit: $readlimit\n";
7252 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
7253 $add_pipe->($cstream);
7259 my $info = PVE
::Storage
::decompressor_info
('vma', $comp);
7260 my $cmd = $info->{decompressor
};
7261 push @$cmd, $readfrom;
7265 my $tmpdir = "/var/tmp/vzdumptmp$$";
7268 # disable interrupts (always do cleanups)
7272 local $SIG{HUP
} = sub { warn "got interrupt - ignored\n"; };
7274 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
7275 POSIX
::mkfifo
($mapfifo, 0600);
7277 my $openfifo = sub { open($fifofh, '>', $mapfifo) or die $! };
7279 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
7284 my $devinfo = {}; # info about drives included in backup
7285 my $virtdev_hash = {}; # info about allocated drives
7287 my $rpcenv = PVE
::RPCEnvironment
::get
();
7289 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7291 # Note: $oldconf is undef if VM does not exist
7292 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7293 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7294 my $new_conf_raw = '';
7298 my $print_devmap = sub {
7299 my $cfgfn = "$tmpdir/qemu-server.conf";
7301 # we can read the config - that is already extracted
7302 my $fh = IO
::File-
>new($cfgfn, "r") ||
7303 die "unable to read qemu-server.conf - $!\n";
7305 my $fwcfgfn = "$tmpdir/qemu-server.fw";
7307 my $pve_firewall_dir = '/etc/pve/firewall';
7308 mkdir $pve_firewall_dir; # make sure the dir exists
7309 PVE
::Tools
::file_copy
($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
7312 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
7314 foreach my $info (values %{$virtdev_hash}) {
7315 my $storeid = $info->{storeid
};
7316 next if defined($storage_limits{$storeid});
7318 my $limit = PVE
::Storage
::get_bandwidth_limit
('restore', [$storeid], $bwlimit) // 0;
7319 print STDERR
"rate limit for storage $storeid: $limit KiB/s\n" if $limit;
7320 $storage_limits{$storeid} = $limit * 1024;
7323 foreach my $devname (keys %$devinfo) {
7324 die "found no device mapping information for device '$devname'\n"
7325 if !$devinfo->{$devname}->{virtdev
};
7328 # create empty/temp config
7330 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\n");
7331 $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
7335 my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
7337 # print restore information to $fifofh
7338 foreach my $virtdev (sort keys %$virtdev_hash) {
7339 my $d = $virtdev_hash->{$virtdev};
7340 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7342 my $storeid = $d->{storeid
};
7343 my $volid = $d->{volid
};
7346 if (my $limit = $storage_limits{$storeid}) {
7347 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
7350 my $write_zeros = 1;
7351 if (PVE
::Storage
::volume_has_feature
($cfg, 'sparseinit', $volid)) {
7355 my $path = PVE
::Storage
::path
($cfg, $volid);
7357 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
7359 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
7362 $fh->seek(0, 0) || die "seek failed - $!\n";
7364 my $cookie = { netcount
=> 0 };
7365 while (defined(my $line = <$fh>)) {
7366 $new_conf_raw .= restore_update_config_line
(
7383 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7384 local $SIG{ALRM
} = sub { die "got timeout\n"; };
7386 $oldtimeout = alarm($timeout);
7393 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
7394 my ($dev_id, $size, $devname) = ($1, $2, $3);
7395 $devinfo->{$devname} = { size
=> $size, dev_id
=> $dev_id };
7396 } elsif ($line =~ m/^CTIME: /) {
7397 # we correctly received the vma config, so we can disable
7398 # the timeout now for disk allocation (set to 10 minutes, so
7399 # that we always timeout if something goes wrong)
7402 print $fifofh "done\n";
7403 my $tmp = $oldtimeout || 0;
7404 $oldtimeout = undef;
7411 print "restore vma archive: $dbg_cmdstring\n";
7412 run_command
($commands, input
=> $input, outfunc
=> $parser, afterfork
=> $openfifo);
7416 alarm($oldtimeout) if $oldtimeout;
7418 $restore_deactivate_volumes->($cfg, $virtdev_hash);
7420 close($fifofh) if $fifofh;
7425 $restore_destroy_volumes->($cfg, $virtdev_hash);
7429 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $opts->{override_conf
});
7430 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7432 eval { rescan
($vmid, 1); };
7435 PVE
::AccessControl
::add_vm_to_pool
($vmid, $opts->{pool
}) if $opts->{pool
};
7438 sub restore_tar_archive
{
7439 my ($archive, $vmid, $user, $opts) = @_;
7441 if (scalar(keys $opts->{override_conf
}->%*) > 0) {
7442 my $keystring = join(' ', keys $opts->{override_conf
}->%*);
7443 die "cannot pass along options ($keystring) when restoring from tar archive\n";
7446 if ($archive ne '-') {
7447 my $firstfile = tar_archive_read_firstfile
($archive);
7448 die "ERROR: file '$archive' does not look like a QemuServer vzdump backup\n"
7449 if $firstfile ne 'qemu-server.conf';
7452 my $storecfg = PVE
::Storage
::config
();
7454 # avoid zombie disks when restoring over an existing VM -> cleanup first
7455 # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
7456 # skiplock=1 because qmrestore has set the 'create' lock itself already
7457 my $vmcfgfn = PVE
::QemuConfig-
>config_file($vmid);
7458 destroy_vm
($storecfg, $vmid, 1, { lock => 'restore' }) if -f
$vmcfgfn;
7460 my $tocmd = "/usr/lib/qemu-server/qmextract";
7462 $tocmd .= " --storage " . PVE
::Tools
::shellquote
($opts->{storage
}) if $opts->{storage
};
7463 $tocmd .= " --pool " . PVE
::Tools
::shellquote
($opts->{pool
}) if $opts->{pool
};
7464 $tocmd .= ' --prealloc' if $opts->{prealloc
};
7465 $tocmd .= ' --info' if $opts->{info
};
7467 # tar option "xf" does not autodetect compression when read from STDIN,
7468 # so we pipe to zcat
7469 my $cmd = "zcat -f|tar xf " . PVE
::Tools
::shellquote
($archive) . " " .
7470 PVE
::Tools
::shellquote
("--to-command=$tocmd");
7472 my $tmpdir = "/var/tmp/vzdumptmp$$";
7475 local $ENV{VZDUMP_TMPDIR
} = $tmpdir;
7476 local $ENV{VZDUMP_VMID
} = $vmid;
7477 local $ENV{VZDUMP_USER
} = $user;
7479 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7480 my $new_conf_raw = '';
7482 # disable interrupts (always do cleanups)
7486 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7494 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7496 if ($archive eq '-') {
7497 print "extracting archive from STDIN\n";
7498 run_command
($cmd, input
=> "<&STDIN");
7500 print "extracting archive '$archive'\n";
7504 return if $opts->{info
};
7508 my $statfile = "$tmpdir/qmrestore.stat";
7509 if (my $fd = IO
::File-
>new($statfile, "r")) {
7510 while (defined (my $line = <$fd>)) {
7511 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
7512 $map->{$1} = $2 if $1;
7514 print STDERR
"unable to parse line in statfile - $line\n";
7520 my $confsrc = "$tmpdir/qemu-server.conf";
7522 my $srcfd = IO
::File-
>new($confsrc, "r") || die "unable to open file '$confsrc'\n";
7524 my $cookie = { netcount
=> 0 };
7525 while (defined (my $line = <$srcfd>)) {
7526 $new_conf_raw .= restore_update_config_line
(
7537 tar_restore_cleanup
($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info
};
7543 PVE
::Tools
::file_set_contents
($conffile, $new_conf_raw);
7545 PVE
::Cluster
::cfs_update
(); # make sure we read new file
7547 eval { rescan
($vmid, 1); };
7551 sub foreach_storage_used_by_vm
{
7552 my ($conf, $func) = @_;
7556 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7557 my ($ds, $drive) = @_;
7558 return if drive_is_cdrom
($drive);
7560 my $volid = $drive->{file
};
7562 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
7563 $sidhash->{$sid} = $sid if $sid;
7566 foreach my $sid (sort keys %$sidhash) {
7571 my $qemu_snap_storage = {
7574 sub do_snapshots_with_qemu
{
7575 my ($storecfg, $volid, $deviceid) = @_;
7577 return if $deviceid =~ m/tpmstate0/;
7579 my $storage_name = PVE
::Storage
::parse_volume_id
($volid);
7580 my $scfg = $storecfg->{ids
}->{$storage_name};
7581 die "could not find storage '$storage_name'\n" if !defined($scfg);
7583 if ($qemu_snap_storage->{$scfg->{type
}} && !$scfg->{krbd
}){
7587 if ($volid =~ m/\.(qcow2|qed)$/){
7594 sub qga_check_running
{
7595 my ($vmid, $nowarn) = @_;
7597 eval { mon_cmd
($vmid, "guest-ping", timeout
=> 3); };
7599 warn "QEMU Guest Agent is not running - $@" if !$nowarn;
7605 sub template_create
{
7606 my ($vmid, $conf, $disk) = @_;
7608 my $storecfg = PVE
::Storage
::config
();
7610 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7611 my ($ds, $drive) = @_;
7613 return if drive_is_cdrom
($drive);
7614 return if $disk && $ds ne $disk;
7616 my $volid = $drive->{file
};
7617 return if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
7619 my $voliddst = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
7620 $drive->{file
} = $voliddst;
7621 $conf->{$ds} = print_drive
($drive);
7622 PVE
::QemuConfig-
>write_config($vmid, $conf);
7626 sub convert_iscsi_path
{
7629 if ($path =~ m
|^iscsi
://([^/]+)/([^/]+)/(.+)$|) {
7634 my $initiator_name = get_initiator_name
();
7636 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
7637 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
7640 die "cannot convert iscsi path '$path', unkown format\n";
7643 sub qemu_img_convert
{
7644 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
7646 my $storecfg = PVE
::Storage
::config
();
7647 my ($src_storeid, $src_volname) = PVE
::Storage
::parse_volume_id
($src_volid, 1);
7648 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid, 1);
7650 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
7654 my $src_is_iscsi = 0;
7658 PVE
::Storage
::activate_volumes
($storecfg, [$src_volid], $snapname);
7659 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7660 $src_format = qemu_img_format
($src_scfg, $src_volname);
7661 $src_path = PVE
::Storage
::path
($storecfg, $src_volid, $snapname);
7662 $src_is_iscsi = ($src_path =~ m
|^iscsi
://|);
7663 $cachemode = 'none' if $src_scfg->{type
} eq 'zfspool';
7664 } elsif (-f
$src_volid || -b
$src_volid) {
7665 $src_path = $src_volid;
7666 if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7671 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
7673 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7674 my $dst_format = qemu_img_format
($dst_scfg, $dst_volname);
7675 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7676 my $dst_is_iscsi = ($dst_path =~ m
|^iscsi
://|);
7679 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
7680 push @$cmd, '-l', "snapshot.name=$snapname"
7681 if $snapname && $src_format && $src_format eq "qcow2";
7682 push @$cmd, '-t', 'none' if $dst_scfg->{type
} eq 'zfspool';
7683 push @$cmd, '-T', $cachemode if defined($cachemode);
7685 if ($src_is_iscsi) {
7686 push @$cmd, '--image-opts';
7687 $src_path = convert_iscsi_path
($src_path);
7688 } elsif ($src_format) {
7689 push @$cmd, '-f', $src_format;
7692 if ($dst_is_iscsi) {
7693 push @$cmd, '--target-image-opts';
7694 $dst_path = convert_iscsi_path
($dst_path);
7696 push @$cmd, '-O', $dst_format;
7699 push @$cmd, $src_path;
7701 if (!$dst_is_iscsi && $is_zero_initialized) {
7702 push @$cmd, "zeroinit:$dst_path";
7704 push @$cmd, $dst_path;
7709 if($line =~ m/\((\S+)\/100\
%\)/){
7711 my $transferred = int($size * $percent / 100);
7712 my $total_h = render_bytes
($size, 1);
7713 my $transferred_h = render_bytes
($transferred, 1);
7715 print "transferred $transferred_h of $total_h ($percent%)\n";
7720 eval { run_command
($cmd, timeout
=> undef, outfunc
=> $parser); };
7722 die "copy failed: $err" if $err;
7725 sub qemu_img_format
{
7726 my ($scfg, $volname) = @_;
7728 if ($scfg->{path
} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7735 sub qemu_drive_mirror
{
7736 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
7738 $jobs = {} if !$jobs;
7742 $jobs->{"drive-$drive"} = {};
7744 if ($dst_volid =~ /^nbd:/) {
7745 $qemu_target = $dst_volid;
7748 my $storecfg = PVE
::Storage
::config
();
7749 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid);
7751 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7753 $format = qemu_img_format
($dst_scfg, $dst_volname);
7755 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7757 $qemu_target = $is_zero_initialized ?
"zeroinit:$dst_path" : $dst_path;
7760 my $opts = { timeout
=> 10, device
=> "drive-$drive", mode
=> "existing", sync
=> "full", target
=> $qemu_target };
7761 $opts->{format
} = $format if $format;
7763 if (defined($src_bitmap)) {
7764 $opts->{sync
} = 'incremental';
7765 $opts->{bitmap
} = $src_bitmap;
7766 print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
7769 if (defined($bwlimit)) {
7770 $opts->{speed
} = $bwlimit * 1024;
7771 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7773 print "drive mirror is starting for drive-$drive\n";
7776 # if a job already runs for this device we get an error, catch it for cleanup
7777 eval { mon_cmd
($vmid, "drive-mirror", %$opts); };
7779 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7781 die "mirroring error: $err\n";
7784 qemu_drive_mirror_monitor
($vmid, $vmiddst, $jobs, $completion, $qga);
7787 # $completion can be either
7788 # 'complete': wait until all jobs are ready, block-job-complete them (default)
7789 # 'cancel': wait until all jobs are ready, block-job-cancel them
7790 # 'skip': wait until all jobs are ready, return with block jobs in ready state
7791 # 'auto': wait until all jobs disappear, only use for jobs which complete automatically
7792 sub qemu_drive_mirror_monitor
{
7793 my ($vmid, $vmiddst, $jobs, $completion, $qga, $op) = @_;
7795 $completion //= 'complete';
7799 my $err_complete = 0;
7801 my $starttime = time ();
7803 die "block job ('$op') timed out\n" if $err_complete > 300;
7805 my $stats = mon_cmd
($vmid, "query-block-jobs");
7808 my $running_jobs = {};
7809 for my $stat (@$stats) {
7810 next if $stat->{type
} ne $op;
7811 $running_jobs->{$stat->{device
}} = $stat;
7814 my $readycounter = 0;
7816 for my $job_id (sort keys %$jobs) {
7817 my $job = $running_jobs->{$job_id};
7819 my $vanished = !defined($job);
7820 my $complete = defined($jobs->{$job_id}->{complete
}) && $vanished;
7821 if($complete || ($vanished && $completion eq 'auto')) {
7822 print "$job_id: $op-job finished\n";
7823 delete $jobs->{$job_id};
7827 die "$job_id: '$op' has been cancelled\n" if !defined($job);
7829 my $busy = $job->{busy
};
7830 my $ready = $job->{ready
};
7831 if (my $total = $job->{len
}) {
7832 my $transferred = $job->{offset
} || 0;
7833 my $remaining = $total - $transferred;
7834 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7836 my $duration = $ctime - $starttime;
7837 my $total_h = render_bytes
($total, 1);
7838 my $transferred_h = render_bytes
($transferred, 1);
7840 my $status = sprintf(
7841 "transferred $transferred_h of $total_h ($percent%%) in %s",
7842 render_duration
($duration),
7847 $status .= ", still busy"; # shouldn't even happen? but mirror is weird
7849 $status .= ", ready";
7852 print "$job_id: $status\n" if !$jobs->{$job_id}->{ready
};
7853 $jobs->{$job_id}->{ready
} = $ready;
7856 $readycounter++ if $job->{ready
};
7859 last if scalar(keys %$jobs) == 0;
7861 if ($readycounter == scalar(keys %$jobs)) {
7862 print "all '$op' jobs are ready\n";
7864 # do the complete later (or has already been done)
7865 last if $completion eq 'skip' || $completion eq 'auto';
7867 if ($vmiddst && $vmiddst != $vmid) {
7868 my $agent_running = $qga && qga_check_running
($vmid);
7869 if ($agent_running) {
7870 print "freeze filesystem\n";
7871 eval { mon_cmd
($vmid, "guest-fsfreeze-freeze"); };
7874 print "suspend vm\n";
7875 eval { PVE
::QemuServer
::vm_suspend
($vmid, 1); };
7879 # if we clone a disk for a new target vm, we don't switch the disk
7880 PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs);
7882 if ($agent_running) {
7883 print "unfreeze filesystem\n";
7884 eval { mon_cmd
($vmid, "guest-fsfreeze-thaw"); };
7887 print "resume vm\n";
7888 eval { PVE
::QemuServer
::vm_resume
($vmid, 1, 1); };
7895 for my $job_id (sort keys %$jobs) {
7896 # try to switch the disk if source and destination are on the same guest
7897 print "$job_id: Completing block job_id...\n";
7900 if ($completion eq 'complete') {
7901 $op = 'block-job-complete';
7902 } elsif ($completion eq 'cancel') {
7903 $op = 'block-job-cancel';
7905 die "invalid completion value: $completion\n";
7907 eval { mon_cmd
($vmid, $op, device
=> $job_id) };
7908 if ($@ =~ m/cannot be completed/) {
7909 print "$job_id: block job cannot be completed, trying again.\n";
7912 print "$job_id: Completed successfully.\n";
7913 $jobs->{$job_id}->{complete
} = 1;
7924 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7925 die "block job ($op) error: $err";
7929 sub qemu_blockjobs_cancel
{
7930 my ($vmid, $jobs) = @_;
7932 foreach my $job (keys %$jobs) {
7933 print "$job: Cancelling block job\n";
7934 eval { mon_cmd
($vmid, "block-job-cancel", device
=> $job); };
7935 $jobs->{$job}->{cancel
} = 1;
7939 my $stats = mon_cmd
($vmid, "query-block-jobs");
7941 my $running_jobs = {};
7942 foreach my $stat (@$stats) {
7943 $running_jobs->{$stat->{device
}} = $stat;
7946 foreach my $job (keys %$jobs) {
7948 if (defined($jobs->{$job}->{cancel
}) && !defined($running_jobs->{$job})) {
7949 print "$job: Done.\n";
7950 delete $jobs->{$job};
7954 last if scalar(keys %$jobs) == 0;
7960 # Check for bug #4525: drive-mirror will open the target drive with the same aio setting as the
7961 # source, but some storages have problems with io_uring, sometimes even leading to crashes.
7962 my sub clone_disk_check_io_uring
{
7963 my ($src_drive, $storecfg, $src_storeid, $dst_storeid, $use_drive_mirror) = @_;
7965 return if !$use_drive_mirror;
7967 # Don't complain when not changing storage.
7968 # Assume if it works for the source, it'll work for the target too.
7969 return if $src_storeid eq $dst_storeid;
7971 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7972 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7974 my $cache_direct = drive_uses_cache_direct
($src_drive);
7976 my $src_uses_io_uring;
7977 if ($src_drive->{aio
}) {
7978 $src_uses_io_uring = $src_drive->{aio
} eq 'io_uring';
7980 $src_uses_io_uring = storage_allows_io_uring_default
($src_scfg, $cache_direct);
7983 die "target storage is known to cause issues with aio=io_uring (used by current drive)\n"
7984 if $src_uses_io_uring && !storage_allows_io_uring_default
($dst_scfg, $cache_direct);
7988 my ($storecfg, $source, $dest, $full, $newvollist, $jobs, $completion, $qga, $bwlimit) = @_;
7990 my ($vmid, $running) = $source->@{qw(vmid running)};
7991 my ($src_drivename, $drive, $snapname) = $source->@{qw(drivename drive snapname)};
7993 my ($newvmid, $dst_drivename, $efisize) = $dest->@{qw(vmid drivename efisize)};
7994 my ($storage, $format) = $dest->@{qw(storage format)};
7996 my $use_drive_mirror = $full && $running && $src_drivename && !$snapname;
7998 if ($src_drivename && $dst_drivename && $src_drivename ne $dst_drivename) {
7999 die "cloning from/to EFI disk requires EFI disk\n"
8000 if $src_drivename eq 'efidisk0' || $dst_drivename eq 'efidisk0';
8001 die "cloning from/to TPM state requires TPM state\n"
8002 if $src_drivename eq 'tpmstate0' || $dst_drivename eq 'tpmstate0';
8004 # This would lead to two device nodes in QEMU pointing to the same backing image!
8005 die "cannot change drive name when cloning disk from/to the same VM\n"
8006 if $use_drive_mirror && $vmid == $newvmid;
8009 die "cannot move TPM state while VM is running\n"
8010 if $use_drive_mirror && $src_drivename eq 'tpmstate0';
8014 print "create " . ($full ?
'full' : 'linked') . " clone of drive ";
8015 print "$src_drivename " if $src_drivename;
8016 print "($drive->{file})\n";
8019 $newvolid = PVE
::Storage
::vdisk_clone
($storecfg, $drive->{file
}, $newvmid, $snapname);
8020 push @$newvollist, $newvolid;
8022 my ($src_storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
8023 my $storeid = $storage || $src_storeid;
8025 my $dst_format = resolve_dst_disk_format
($storecfg, $storeid, $volname, $format);
8029 if (drive_is_cloudinit
($drive)) {
8030 $name = "vm-$newvmid-cloudinit";
8031 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8032 if ($scfg->{path
}) {
8033 $name .= ".$dst_format";
8036 $size = PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
;
8037 } elsif ($dst_drivename eq 'efidisk0') {
8038 $size = $efisize or die "internal error - need to specify EFI disk size\n";
8039 } elsif ($dst_drivename eq 'tpmstate0') {
8040 $dst_format = 'raw';
8041 $size = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8043 clone_disk_check_io_uring
($drive, $storecfg, $src_storeid, $storeid, $use_drive_mirror);
8045 ($size) = PVE
::Storage
::volume_size_info
($storecfg, $drive->{file
}, 10);
8047 $newvolid = PVE
::Storage
::vdisk_alloc
(
8048 $storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024)
8050 push @$newvollist, $newvolid;
8052 PVE
::Storage
::activate_volumes
($storecfg, [$newvolid]);
8054 if (drive_is_cloudinit
($drive)) {
8055 # when cloning multiple disks (e.g. during clone_vm) it might be the last disk
8056 # if this is the case, we have to complete any block-jobs still there from
8057 # previous drive-mirrors
8058 if (($completion eq 'complete') && (scalar(keys %$jobs) > 0)) {
8059 qemu_drive_mirror_monitor
($vmid, $newvmid, $jobs, $completion, $qga);
8064 my $sparseinit = PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $newvolid);
8065 if ($use_drive_mirror) {
8066 qemu_drive_mirror
($vmid, $src_drivename, $newvolid, $newvmid, $sparseinit, $jobs,
8067 $completion, $qga, $bwlimit);
8069 # TODO: handle bwlimits
8070 if ($dst_drivename eq 'efidisk0') {
8071 # the relevant data on the efidisk may be smaller than the source
8072 # e.g. on RBD/ZFS, so we use dd to copy only the amount
8073 # that is given by the OVMF_VARS.fd
8074 my $src_path = PVE
::Storage
::path
($storecfg, $drive->{file
}, $snapname);
8075 my $dst_path = PVE
::Storage
::path
($storecfg, $newvolid);
8077 my $src_format = (PVE
::Storage
::parse_volname
($storecfg, $drive->{file
}))[6];
8079 # better for Ceph if block size is not too small, see bug #3324
8082 my $cmd = ['qemu-img', 'dd', '-n', '-O', $dst_format];
8084 if ($src_format eq 'qcow2' && $snapname) {
8085 die "cannot clone qcow2 EFI disk snapshot - requires QEMU >= 6.2\n"
8086 if !min_version
(kvm_user_version
(), 6, 2);
8087 push $cmd->@*, '-l', $snapname;
8089 push $cmd->@*, "bs=$bs", "osize=$size", "if=$src_path", "of=$dst_path";
8092 qemu_img_convert
($drive->{file
}, $newvolid, $size, $snapname, $sparseinit);
8098 my ($size) = eval { PVE
::Storage
::volume_size_info
($storecfg, $newvolid, 10) };
8100 my $disk = dclone
($drive);
8101 delete $disk->{format
};
8102 $disk->{file
} = $newvolid;
8103 $disk->{size
} = $size if defined($size);
8108 sub get_running_qemu_version
{
8110 my $res = mon_cmd
($vmid, "query-version");
8111 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
8114 sub qemu_use_old_bios_files
{
8115 my ($machine_type) = @_;
8117 return if !$machine_type;
8119 my $use_old_bios_files = undef;
8121 if ($machine_type =~ m/^(\S+)\.pxe$/) {
8123 $use_old_bios_files = 1;
8125 my $version = extract_version
($machine_type, kvm_user_version
());
8126 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
8127 # load new efi bios files on migration. So this hack is required to allow
8128 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
8129 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
8130 $use_old_bios_files = !min_version
($version, 2, 4);
8133 return ($use_old_bios_files, $machine_type);
8136 sub get_efivars_size
{
8137 my ($conf, $efidisk) = @_;
8139 my $arch = get_vm_arch
($conf);
8140 $efidisk //= $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
8141 my $smm = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
8142 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8143 return -s
$ovmf_vars;
8146 sub update_efidisk_size
{
8149 return if !defined($conf->{efidisk0
});
8151 my $disk = PVE
::QemuServer
::parse_drive
('efidisk0', $conf->{efidisk0
});
8152 $disk->{size
} = get_efivars_size
($conf);
8153 $conf->{efidisk0
} = print_drive
($disk);
8158 sub update_tpmstate_size
{
8161 my $disk = PVE
::QemuServer
::parse_drive
('tpmstate0', $conf->{tpmstate0
});
8162 $disk->{size
} = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8163 $conf->{tpmstate0
} = print_drive
($disk);
8166 sub create_efidisk
($$$$$$$) {
8167 my ($storecfg, $storeid, $vmid, $fmt, $arch, $efidisk, $smm) = @_;
8169 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8171 my $vars_size_b = -s
$ovmf_vars;
8172 my $vars_size = PVE
::Tools
::convert_size
($vars_size_b, 'b' => 'kb');
8173 my $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
8174 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
8176 qemu_img_convert
($ovmf_vars, $volid, $vars_size_b, undef, 0);
8177 my ($size) = PVE
::Storage
::volume_size_info
($storecfg, $volid, 3);
8179 return ($volid, $size/1024);
8182 sub vm_iothreads_list
{
8185 my $res = mon_cmd
($vmid, 'query-iothreads');
8188 foreach my $iothread (@$res) {
8189 $iothreads->{ $iothread->{id
} } = $iothread->{"thread-id"};
8196 my ($conf, $drive) = @_;
8200 if (!$conf->{scsihw
} || ($conf->{scsihw
} =~ m/^lsi/)) {
8202 } elsif ($conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single')) {
8208 my $controller = int($drive->{index} / $maxdev);
8209 my $controller_prefix = ($conf->{scsihw
} && $conf->{scsihw
} eq 'virtio-scsi-single')
8213 return ($maxdev, $controller, $controller_prefix);
8216 sub resolve_dst_disk_format
{
8217 my ($storecfg, $storeid, $src_volname, $format) = @_;
8218 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
8221 # if no target format is specified, use the source disk format as hint
8223 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8224 $format = qemu_img_format
($scfg, $src_volname);
8230 # test if requested format is supported - else use default
8231 my $supported = grep { $_ eq $format } @$validFormats;
8232 $format = $defFormat if !$supported;
8236 # NOTE: if this logic changes, please update docs & possibly gui logic
8237 sub find_vmstate_storage
{
8238 my ($conf, $storecfg) = @_;
8240 # first, return storage from conf if set
8241 return $conf->{vmstatestorage
} if $conf->{vmstatestorage
};
8243 my ($target, $shared, $local);
8245 foreach_storage_used_by_vm
($conf, sub {
8247 my $scfg = PVE
::Storage
::storage_config
($storecfg, $sid);
8248 my $dst = $scfg->{shared
} ? \
$shared : \
$local;
8249 $$dst = $sid if !$$dst || $scfg->{path
}; # prefer file based storage
8252 # second, use shared storage where VM has at least one disk
8253 # third, use local storage where VM has at least one disk
8254 # fall back to local storage
8255 $target = $shared // $local // 'local';
8261 my ($uuid, $uuid_str);
8262 UUID
::generate
($uuid);
8263 UUID
::unparse
($uuid, $uuid_str);
8267 sub generate_smbios1_uuid
{
8268 return "uuid=".generate_uuid
();
8274 mon_cmd
($vmid, 'nbd-server-stop');
8277 sub create_reboot_request
{
8279 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
8280 or die "failed to create reboot trigger file: $!\n";
8284 sub clear_reboot_request
{
8286 my $path = "/run/qemu-server/$vmid.reboot";
8289 $res = unlink($path);
8290 die "could not remove reboot request for $vmid: $!"
8291 if !$res && $! != POSIX
::ENOENT
;
8296 sub bootorder_from_legacy
{
8297 my ($conf, $bootcfg) = @_;
8299 my $boot = $bootcfg->{legacy
} || $boot_fmt->{legacy
}->{default};
8300 my $bootindex_hash = {};
8302 foreach my $o (split(//, $boot)) {
8303 $bootindex_hash->{$o} = $i*100;
8309 PVE
::QemuConfig-
>foreach_volume($conf, sub {
8310 my ($ds, $drive) = @_;
8312 if (drive_is_cdrom
($drive, 1)) {
8313 if ($bootindex_hash->{d
}) {
8314 $bootorder->{$ds} = $bootindex_hash->{d
};
8315 $bootindex_hash->{d
} += 1;
8317 } elsif ($bootindex_hash->{c
}) {
8318 $bootorder->{$ds} = $bootindex_hash->{c
}
8319 if $conf->{bootdisk
} && $conf->{bootdisk
} eq $ds;
8320 $bootindex_hash->{c
} += 1;
8324 if ($bootindex_hash->{n
}) {
8325 for (my $i = 0; $i < $MAX_NETS; $i++) {
8326 my $netname = "net$i";
8327 next if !$conf->{$netname};
8328 $bootorder->{$netname} = $bootindex_hash->{n
};
8329 $bootindex_hash->{n
} += 1;
8336 # Generate default device list for 'boot: order=' property. Matches legacy
8337 # default boot order, but with explicit device names. This is important, since
8338 # the fallback for when neither 'order' nor the old format is specified relies
8339 # on 'bootorder_from_legacy' above, and it would be confusing if this diverges.
8340 sub get_default_bootdevices
{
8346 my $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 0);
8347 push @ret, $first if $first;
8350 $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 1);
8351 push @ret, $first if $first;
8354 for (my $i = 0; $i < $MAX_NETS; $i++) {
8355 my $netname = "net$i";
8356 next if !$conf->{$netname};
8357 push @ret, $netname;
8364 sub device_bootorder
{
8367 return bootorder_from_legacy
($conf) if !defined($conf->{boot
});
8369 my $boot = parse_property_string
($boot_fmt, $conf->{boot
});
8372 if (!defined($boot) || $boot->{legacy
}) {
8373 $bootorder = bootorder_from_legacy
($conf, $boot);
8374 } elsif ($boot->{order
}) {
8375 my $i = 100; # start at 100 to allow user to insert devices before us with -args
8376 for my $dev (PVE
::Tools
::split_list
($boot->{order
})) {
8377 $bootorder->{$dev} = $i++;
8384 sub register_qmeventd_handle
{
8388 my $peer = "/var/run/qmeventd.sock";
8393 $fh = IO
::Socket
::UNIX-
>new(Peer
=> $peer, Blocking
=> 0, Timeout
=> 1);
8395 if ($! != EINTR
&& $! != EAGAIN
) {
8396 die "unable to connect to qmeventd socket (vmid: $vmid) - $!\n";
8399 die "unable to connect to qmeventd socket (vmid: $vmid) - timeout "
8400 . "after $count retries\n";
8405 # send handshake to mark VM as backing up
8406 print $fh to_json
({vzdump
=> {vmid
=> "$vmid"}});
8408 # return handle to be closed later when inhibit is no longer required
8412 # bash completion helper
8414 sub complete_backup_archives
{
8415 my ($cmdname, $pname, $cvalue) = @_;
8417 my $cfg = PVE
::Storage
::config
();
8421 if ($cvalue =~ m/^([^:]+):/) {
8425 my $data = PVE
::Storage
::template_list
($cfg, $storeid, 'backup');
8428 foreach my $id (keys %$data) {
8429 foreach my $item (@{$data->{$id}}) {
8430 next if $item->{format
} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
8431 push @$res, $item->{volid
} if defined($item->{volid
});
8438 my $complete_vmid_full = sub {
8441 my $idlist = vmstatus
();
8445 foreach my $id (keys %$idlist) {
8446 my $d = $idlist->{$id};
8447 if (defined($running)) {
8448 next if $d->{template
};
8449 next if $running && $d->{status
} ne 'running';
8450 next if !$running && $d->{status
} eq 'running';
8459 return &$complete_vmid_full();
8462 sub complete_vmid_stopped
{
8463 return &$complete_vmid_full(0);
8466 sub complete_vmid_running
{
8467 return &$complete_vmid_full(1);
8470 sub complete_storage
{
8472 my $cfg = PVE
::Storage
::config
();
8473 my $ids = $cfg->{ids
};
8476 foreach my $sid (keys %$ids) {
8477 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, undef, 1);
8478 next if !$ids->{$sid}->{content
}->{images
};
8485 sub complete_migration_storage
{
8486 my ($cmd, $param, $current_value, $all_args) = @_;
8488 my $targetnode = @$all_args[1];
8490 my $cfg = PVE
::Storage
::config
();
8491 my $ids = $cfg->{ids
};
8494 foreach my $sid (keys %$ids) {
8495 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, $targetnode, 1);
8496 next if !$ids->{$sid}->{content
}->{images
};
8505 my $qmpstatus = eval {
8506 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid);
8507 mon_cmd
($vmid, "query-status");
8510 return $qmpstatus && $qmpstatus->{status
} eq "paused";
8513 sub check_volume_storage_type
{
8514 my ($storecfg, $vol) = @_;
8516 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($vol);
8517 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8518 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $vol);
8520 die "storage '$storeid' does not support content-type '$vtype'\n"
8521 if !$scfg->{content
}->{$vtype};
8526 sub add_nets_bridge_fdb
{
8527 my ($conf, $vmid) = @_;
8529 for my $opt (keys %$conf) {
8530 next if $opt !~ m/^net(\d+)$/;
8531 my $iface = "tap${vmid}i$1";
8532 # NOTE: expect setups with learning off to *not* use auto-random-generation of MAC on start
8533 my $net = parse_net
($conf->{$opt}, 1) or next;
8535 my $mac = $net->{macaddr
};
8537 log_warn
("MAC learning disabled, but vNIC '$iface' has no static MAC to add to forwarding DB!")
8538 if !file_read_firstline
("/sys/class/net/$iface/brport/learning");
8542 my $bridge = $net->{bridge
};
8544 PVE
::Network
::SDN
::Zones
::add_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8545 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8546 PVE
::Network
::add_bridge_fdb
($iface, $mac, $net->{firewall
});
8551 sub del_nets_bridge_fdb
{
8552 my ($conf, $vmid) = @_;
8554 for my $opt (keys %$conf) {
8555 next if $opt !~ m/^net(\d+)$/;
8556 my $iface = "tap${vmid}i$1";
8558 my $net = parse_net
($conf->{$opt}) or next;
8559 my $mac = $net->{macaddr
} or next;
8561 my $bridge = $net->{bridge
};
8563 PVE
::Network
::SDN
::Zones
::del_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8564 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8565 PVE
::Network
::del_bridge_fdb
($iface, $mac, $net->{firewall
});