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
(
3888 $conf, $vmid, $sockets, $cores, $defaults, $hotplug_features->{memory
}, $cmd);
3890 push @$cmd, '-S' if $conf->{freeze
};
3892 push @$cmd, '-k', $conf->{keyboard
} if defined($conf->{keyboard
});
3894 my $guest_agent = parse_guest_agent
($conf);
3896 if ($guest_agent->{enabled
}) {
3897 my $qgasocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid, 1);
3898 push @$devices, '-chardev', "socket,path=$qgasocket,server=on,wait=off,id=qga0";
3900 if (!$guest_agent->{type
} || $guest_agent->{type
} eq 'virtio') {
3901 my $pciaddr = print_pci_addr
("qga0", $bridges, $arch, $machine_type);
3902 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3903 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3904 } elsif ($guest_agent->{type
} eq 'isa') {
3905 push @$devices, '-device', "isa-serial,chardev=qga0";
3909 my $rng = $conf->{rng0
} ? parse_rng
($conf->{rng0
}) : undef;
3910 if ($rng && $version_guard->(4, 1, 2)) {
3911 check_rng_source
($rng->{source
});
3913 my $max_bytes = $rng->{max_bytes
} // $rng_fmt->{max_bytes
}->{default};
3914 my $period = $rng->{period
} // $rng_fmt->{period
}->{default};
3915 my $limiter_str = "";
3917 $limiter_str = ",max-bytes=$max_bytes,period=$period";
3920 my $rng_addr = print_pci_addr
("rng0", $bridges, $arch, $machine_type);
3921 push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3922 push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3927 if ($qxlnum || $vga->{type
} =~ /^virtio/) {
3930 for (my $i = 1; $i < $qxlnum; $i++){
3931 push @$devices, '-device', print_vga_device
(
3932 $conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3935 # assume other OS works like Linux
3936 my ($ram, $vram) = ("134217728", "67108864");
3937 if ($vga->{memory
}) {
3938 $ram = PVE
::Tools
::convert_size
($qxlnum*4*$vga->{memory
}, 'mb' => 'b');
3939 $vram = PVE
::Tools
::convert_size
($qxlnum*2*$vga->{memory
}, 'mb' => 'b');
3941 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3942 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3946 my $pciaddr = print_pci_addr
("spice", $bridges, $arch, $machine_type);
3948 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
3949 my @nodeaddrs = PVE
::Tools
::getaddrinfo_all
('localhost', family
=> $pfamily);
3950 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3952 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3953 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3954 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3956 my $localhost = PVE
::Network
::addr_to_ip
($nodeaddrs[0]->{addr
});
3957 $spice_port = PVE
::Tools
::next_spice_port
($pfamily, $localhost);
3959 my $spice_enhancement_str = $conf->{spice_enhancements
} // '';
3960 my $spice_enhancement = parse_property_string
($spice_enhancements_fmt, $spice_enhancement_str);
3961 if ($spice_enhancement->{foldersharing
}) {
3962 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3963 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3966 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3967 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}"
3968 if $spice_enhancement->{videostreaming
};
3970 push @$devices, '-spice', "$spice_opts";
3973 # enable balloon by default, unless explicitly disabled
3974 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
3975 my $pciaddr = print_pci_addr
("balloon0", $bridges, $arch, $machine_type);
3976 my $ballooncmd = "virtio-balloon-pci,id=balloon0$pciaddr";
3977 $ballooncmd .= ",free-page-reporting=on" if min_version
($machine_version, 6, 2);
3978 push @$devices, '-device', $ballooncmd;
3981 if ($conf->{watchdog
}) {
3982 my $wdopts = parse_watchdog
($conf->{watchdog
});
3983 my $pciaddr = print_pci_addr
("watchdog", $bridges, $arch, $machine_type);
3984 my $watchdog = $wdopts->{model
} || 'i6300esb';
3985 push @$devices, '-device', "$watchdog$pciaddr";
3986 push @$devices, '-watchdog-action', $wdopts->{action
} if $wdopts->{action
};
3990 my $scsicontroller = {};
3991 my $ahcicontroller = {};
3992 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : $defaults->{scsihw
};
3994 # Add iscsi initiator name if available
3995 if (my $initiator = get_initiator_name
()) {
3996 push @$devices, '-iscsi', "initiator-name=$initiator";
3999 PVE
::QemuConfig-
>foreach_volume($conf, sub {
4000 my ($ds, $drive) = @_;
4002 if (PVE
::Storage
::parse_volume_id
($drive->{file
}, 1)) {
4003 check_volume_storage_type
($storecfg, $drive->{file
});
4004 push @$vollist, $drive->{file
};
4007 # ignore efidisk here, already added in bios/fw handling code above
4008 return if $drive->{interface
} eq 'efidisk';
4010 return if $drive->{interface
} eq 'tpmstate';
4012 $use_virtio = 1 if $ds =~ m/^virtio/;
4014 $drive->{bootindex
} = $bootorder->{$ds} if $bootorder->{$ds};
4016 if ($drive->{interface
} eq 'virtio'){
4017 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread
};
4020 if ($drive->{interface
} eq 'scsi') {
4022 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
4024 die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
4025 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
4027 my $pciaddr = print_pci_addr
("$controller_prefix$controller", $bridges, $arch, $machine_type);
4028 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ?
"virtio-scsi-pci" : $scsihw;
4031 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{iothread
}){
4032 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4033 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4034 } elsif ($drive->{iothread
}) {
4036 "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n"
4041 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{queues
}){
4042 $queues = ",num_queues=$drive->{queues}";
4045 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues"
4046 if !$scsicontroller->{$controller};
4047 $scsicontroller->{$controller}=1;
4050 if ($drive->{interface
} eq 'sata') {
4051 my $controller = int($drive->{index} / $PVE::QemuServer
::Drive
::MAX_SATA_DISKS
);
4052 my $pciaddr = print_pci_addr
("ahci$controller", $bridges, $arch, $machine_type);
4053 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr"
4054 if !$ahcicontroller->{$controller};
4055 $ahcicontroller->{$controller}=1;
4058 my $pbs_conf = $pbs_backing->{$ds};
4059 my $pbs_name = undef;
4061 $pbs_name = "drive-$ds-pbs";
4062 push @$devices, '-blockdev', print_pbs_blockdev
($pbs_conf, $pbs_name);
4065 my $drive_cmd = print_drive_commandline_full
(
4066 $storecfg, $vmid, $drive, $pbs_name, min_version
($kvmver, 6, 0));
4068 # extra protection for templates, but SATA and IDE don't support it..
4069 $drive_cmd .= ',readonly=on' if drive_is_read_only
($conf, $drive);
4071 push @$devices, '-drive',$drive_cmd;
4072 push @$devices, '-device', print_drivedevice_full
(
4073 $storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4076 for (my $i = 0; $i < $MAX_NETS; $i++) {
4077 my $netname = "net$i";
4079 next if !$conf->{$netname};
4080 my $d = parse_net
($conf->{$netname});
4082 # save the MAC addr here (could be auto-gen. in some odd setups) for FDB registering later?
4084 $use_virtio = 1 if $d->{model
} eq 'virtio';
4086 $d->{bootindex
} = $bootorder->{$netname} if $bootorder->{$netname};
4088 my $netdevfull = print_netdev_full
($vmid, $conf, $arch, $d, $netname);
4089 push @$devices, '-netdev', $netdevfull;
4091 my $netdevicefull = print_netdevice_full
(
4092 $vmid, $conf, $d, $netname, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version);
4094 push @$devices, '-device', $netdevicefull;
4097 if ($conf->{ivshmem
}) {
4098 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
4102 $bus = print_pcie_addr
("ivshmem");
4104 $bus = print_pci_addr
("ivshmem", $bridges, $arch, $machine_type);
4107 my $ivshmem_name = $ivshmem->{name
} // $vmid;
4108 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4110 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4111 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path"
4112 .",size=$ivshmem->{size}M";
4115 # pci.4 is nested in pci.1
4116 $bridges->{1} = 1 if $bridges->{4};
4118 if (!$q35) { # add pci bridges
4119 if (min_version
($machine_version, 2, 3)) {
4123 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4126 for my $k (sort {$b cmp $a} keys %$bridges) {
4127 next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
4130 if ($k == 2 && $legacy_igd) {
4133 my $pciaddr = print_pci_addr
("pci.$k_name", undef, $arch, $machine_type);
4134 my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
4136 if ($q35) { # add after -readconfig pve-q35.cfg
4137 splice @$devices, 2, 0, '-device', $devstr;
4139 unshift @$devices, '-device', $devstr if $k > 0;
4144 push @$machineFlags, 'accel=tcg';
4147 push @$machineFlags, 'smm=off' if should_disable_smm
($conf, $vga);
4149 my $machine_type_min = $machine_type;
4150 if ($add_pve_version) {
4151 $machine_type_min =~ s/\+pve\d+$//;
4152 $machine_type_min .= "+pve$required_pve_version";
4154 push @$machineFlags, "type=${machine_type_min}";
4156 push @$cmd, @$devices;
4157 push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
4158 push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
4159 push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
4161 if (my $vmstate = $conf->{vmstate
}) {
4162 my $statepath = PVE
::Storage
::path
($storecfg, $vmstate);
4163 push @$vollist, $vmstate;
4164 push @$cmd, '-loadstate', $statepath;
4165 print "activating and using '$vmstate' as vmstate\n";
4168 if (PVE
::QemuConfig-
>is_template($conf)) {
4169 # needed to workaround base volumes being read-only
4170 push @$cmd, '-snapshot';
4174 if ($conf->{args
}) {
4175 my $aa = PVE
::Tools
::split_args
($conf->{args
});
4179 return wantarray ?
($cmd, $vollist, $spice_port) : $cmd;
4182 sub check_rng_source
{
4185 # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
4186 die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
4189 my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
4190 if ($source eq '/dev/hwrng' && file_read_firstline
($rng_current) eq 'none') {
4191 # Needs to abort, otherwise QEMU crashes on first rng access. Note that rng_current cannot
4192 # be changed to 'none' manually, so once the VM is past this point, it's no longer an issue.
4193 die "Cannot start VM with passed-through RNG device: '/dev/hwrng' exists, but"
4194 ." '$rng_current' is set to 'none'. Ensure that a compatible hardware-RNG is attached"
4202 my $res = mon_cmd
($vmid, 'query-spice');
4204 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4207 sub vm_devices_list
{
4210 my $res = mon_cmd
($vmid, 'query-pci');
4211 my $devices_to_check = [];
4213 foreach my $pcibus (@$res) {
4214 push @$devices_to_check, @{$pcibus->{devices
}},
4217 while (@$devices_to_check) {
4219 for my $d (@$devices_to_check) {
4220 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4221 next if !$d->{'pci_bridge'} || !$d->{'pci_bridge'}->{devices
};
4223 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices
}});
4224 push @$to_check, @{$d->{'pci_bridge'}->{devices
}};
4226 $devices_to_check = $to_check;
4229 my $resblock = mon_cmd
($vmid, 'query-block');
4230 foreach my $block (@$resblock) {
4231 if($block->{device
} =~ m/^drive-(\S+)/){
4236 my $resmice = mon_cmd
($vmid, 'query-mice');
4237 foreach my $mice (@$resmice) {
4238 if ($mice->{name
} eq 'QEMU HID Tablet') {
4239 $devices->{tablet
} = 1;
4244 # for usb devices there is no query-usb
4245 # but we can iterate over the entries in
4246 # qom-list path=/machine/peripheral
4247 my $resperipheral = mon_cmd
($vmid, 'qom-list', path
=> '/machine/peripheral');
4248 foreach my $per (@$resperipheral) {
4249 if ($per->{name
} =~ m/^usb(?:redirdev)?\d+$/) {
4250 $devices->{$per->{name
}} = 1;
4258 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4260 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
4262 my $devices_list = vm_devices_list
($vmid);
4263 return 1 if defined($devices_list->{$deviceid});
4265 # add PCI bridge if we need it for the device
4266 qemu_add_pci_bridge
($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type);
4268 if ($deviceid eq 'tablet') {
4269 qemu_deviceadd
($vmid, print_tabletdevice_full
($conf, $arch));
4270 } elsif ($deviceid eq 'keyboard') {
4271 qemu_deviceadd
($vmid, print_keyboarddevice_full
($conf, $arch));
4272 } elsif ($deviceid =~ m/^usbredirdev(\d+)$/) {
4274 qemu_spice_usbredir_chardev_add
($vmid, "usbredirchardev$id");
4275 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_spice_usbdevice
($id, "xhci", $id + 1));
4276 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4277 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_usbdevice_full
($conf, $deviceid, $device, {}, $1 + 1));
4278 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4279 qemu_iothread_add
($vmid, $deviceid, $device);
4281 qemu_driveadd
($storecfg, $vmid, $device);
4282 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4284 qemu_deviceadd
($vmid, $devicefull);
4285 eval { qemu_deviceaddverify
($vmid, $deviceid); };
4287 eval { qemu_drivedel
($vmid, $deviceid); };
4291 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4292 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : "lsi";
4293 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4294 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ?
"virtio-scsi-pci" : $scsihw;
4296 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4298 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread
}) {
4299 qemu_iothread_add
($vmid, $deviceid, $device);
4300 $devicefull .= ",iothread=iothread-$deviceid";
4303 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues
}) {
4304 $devicefull .= ",num_queues=$device->{queues}";
4307 qemu_deviceadd
($vmid, $devicefull);
4308 qemu_deviceaddverify
($vmid, $deviceid);
4309 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4310 qemu_findorcreatescsihw
($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4311 qemu_driveadd
($storecfg, $vmid, $device);
4313 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4314 eval { qemu_deviceadd
($vmid, $devicefull); };
4316 eval { qemu_drivedel
($vmid, $deviceid); };
4320 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4321 return if !qemu_netdevadd
($vmid, $conf, $arch, $device, $deviceid);
4323 my $machine_type = PVE
::QemuServer
::Machine
::qemu_machine_pxe
($vmid, $conf);
4324 my $machine_version = PVE
::QemuServer
::Machine
::extract_version
($machine_type);
4325 my $use_old_bios_files = undef;
4326 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
4328 my $netdevicefull = print_netdevice_full
(
4329 $vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type, $machine_version);
4330 qemu_deviceadd
($vmid, $netdevicefull);
4332 qemu_deviceaddverify
($vmid, $deviceid);
4333 qemu_set_link_status
($vmid, $deviceid, !$device->{link_down
});
4336 eval { qemu_netdevdel
($vmid, $deviceid); };
4340 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4342 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4343 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4345 qemu_deviceadd
($vmid, $devicefull);
4346 qemu_deviceaddverify
($vmid, $deviceid);
4348 die "can't hotplug device '$deviceid'\n";
4354 # fixme: this should raise exceptions on error!
4355 sub vm_deviceunplug
{
4356 my ($vmid, $conf, $deviceid) = @_;
4358 my $devices_list = vm_devices_list
($vmid);
4359 return 1 if !defined($devices_list->{$deviceid});
4361 my $bootdisks = PVE
::QemuServer
::Drive
::get_bootdisks
($conf);
4362 die "can't unplug bootdisk '$deviceid'\n" if grep {$_ eq $deviceid} @$bootdisks;
4364 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard' || $deviceid eq 'xhci') {
4365 qemu_devicedel
($vmid, $deviceid);
4366 } elsif ($deviceid =~ m/^usbredirdev\d+$/) {
4367 qemu_devicedel
($vmid, $deviceid);
4368 qemu_devicedelverify
($vmid, $deviceid);
4369 } elsif ($deviceid =~ m/^usb\d+$/) {
4370 qemu_devicedel
($vmid, $deviceid);
4371 qemu_devicedelverify
($vmid, $deviceid);
4372 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4373 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4375 qemu_devicedel
($vmid, $deviceid);
4376 qemu_devicedelverify
($vmid, $deviceid);
4377 qemu_drivedel
($vmid, $deviceid);
4378 qemu_iothread_del
($vmid, $deviceid, $device);
4379 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4380 qemu_devicedel
($vmid, $deviceid);
4381 qemu_devicedelverify
($vmid, $deviceid);
4382 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4383 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4385 qemu_devicedel
($vmid, $deviceid);
4386 qemu_devicedelverify
($vmid, $deviceid);
4387 qemu_drivedel
($vmid, $deviceid);
4388 qemu_deletescsihw
($conf, $vmid, $deviceid);
4390 qemu_iothread_del
($vmid, "virtioscsi$device->{index}", $device)
4391 if $conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single');
4392 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4393 qemu_devicedel
($vmid, $deviceid);
4394 qemu_devicedelverify
($vmid, $deviceid);
4395 qemu_netdevdel
($vmid, $deviceid);
4397 die "can't unplug device '$deviceid'\n";
4403 sub qemu_spice_usbredir_chardev_add
{
4404 my ($vmid, $id) = @_;
4406 mon_cmd
($vmid, "chardev-add" , (
4417 sub qemu_deviceadd
{
4418 my ($vmid, $devicefull) = @_;
4420 $devicefull = "driver=".$devicefull;
4421 my %options = split(/[=,]/, $devicefull);
4423 mon_cmd
($vmid, "device_add" , %options);
4426 sub qemu_devicedel
{
4427 my ($vmid, $deviceid) = @_;
4429 my $ret = mon_cmd
($vmid, "device_del", id
=> $deviceid);
4432 sub qemu_iothread_add
{
4433 my ($vmid, $deviceid, $device) = @_;
4435 if ($device->{iothread
}) {
4436 my $iothreads = vm_iothreads_list
($vmid);
4437 qemu_objectadd
($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4441 sub qemu_iothread_del
{
4442 my ($vmid, $deviceid, $device) = @_;
4444 if ($device->{iothread
}) {
4445 my $iothreads = vm_iothreads_list
($vmid);
4446 qemu_objectdel
($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4450 sub qemu_objectadd
{
4451 my ($vmid, $objectid, $qomtype) = @_;
4453 mon_cmd
($vmid, "object-add", id
=> $objectid, "qom-type" => $qomtype);
4458 sub qemu_objectdel
{
4459 my ($vmid, $objectid) = @_;
4461 mon_cmd
($vmid, "object-del", id
=> $objectid);
4467 my ($storecfg, $vmid, $device) = @_;
4469 my $kvmver = get_running_qemu_version
($vmid);
4470 my $io_uring = min_version
($kvmver, 6, 0);
4471 my $drive = print_drive_commandline_full
($storecfg, $vmid, $device, undef, $io_uring);
4472 $drive =~ s/\\/\\\\/g;
4473 my $ret = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "drive_add auto \"$drive\"");
4475 # If the command succeeds qemu prints: "OK
"
4476 return 1 if $ret =~ m/OK/s;
4478 die "adding drive failed
: $ret\n";
4482 my ($vmid, $deviceid) = @_;
4484 my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-
$deviceid");
4487 return 1 if $ret eq "";
4489 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4490 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4492 die "deleting drive
$deviceid failed
: $ret\n";
4495 sub qemu_deviceaddverify {
4496 my ($vmid, $deviceid) = @_;
4498 for (my $i = 0; $i <= 5; $i++) {
4499 my $devices_list = vm_devices_list($vmid);
4500 return 1 if defined($devices_list->{$deviceid});
4504 die "error on hotplug device
'$deviceid'\n";
4508 sub qemu_devicedelverify {
4509 my ($vmid, $deviceid) = @_;
4511 # need to verify that the device is correctly removed as device_del
4512 # is async and empty return is not reliable
4514 for (my $i = 0; $i <= 5; $i++) {
4515 my $devices_list = vm_devices_list($vmid);
4516 return 1 if !defined($devices_list->{$deviceid});
4520 die "error on hot-unplugging device
'$deviceid'\n";
4523 sub qemu_findorcreatescsihw {
4524 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4526 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4528 my $scsihwid="$controller_prefix$controller";
4529 my $devices_list = vm_devices_list($vmid);
4531 if (!defined($devices_list->{$scsihwid})) {
4532 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4538 sub qemu_deletescsihw {
4539 my ($conf, $vmid, $opt) = @_;
4541 my $device = parse_drive($opt, $conf->{$opt});
4543 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4544 vm_deviceunplug($vmid, $conf, "virtioscsi
$device->{index}");
4548 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4550 my $devices_list = vm_devices_list($vmid);
4551 foreach my $opt (keys %{$devices_list}) {
4552 if (is_valid_drivename($opt)) {
4553 my $drive = parse_drive($opt, $conf->{$opt});
4554 if ($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4560 my $scsihwid="scsihw
$controller";
4562 vm_deviceunplug($vmid, $conf, $scsihwid);
4567 sub qemu_add_pci_bridge {
4568 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4574 print_pci_addr($device, $bridges, $arch, $machine_type);
4576 while (my ($k, $v) = each %$bridges) {
4579 return 1 if !defined($bridgeid) || $bridgeid < 1;
4581 my $bridge = "pci
.$bridgeid";
4582 my $devices_list = vm_devices_list($vmid);
4584 if (!defined($devices_list->{$bridge})) {
4585 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4591 sub qemu_set_link_status {
4592 my ($vmid, $device, $up) = @_;
4594 mon_cmd($vmid, "set_link
", name => $device,
4595 up => $up ? JSON::true : JSON::false);
4598 sub qemu_netdevadd {
4599 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4601 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4602 my %options = split(/[=,]/, $netdev);
4604 if (defined(my $vhost = $options{vhost})) {
4605 $options{vhost} = JSON::boolean(PVE::JSONSchema::parse_boolean($vhost));
4608 if (defined(my $queues = $options{queues})) {
4609 $options{queues} = $queues + 0;
4612 mon_cmd($vmid, "netdev_add
", %options);
4616 sub qemu_netdevdel {
4617 my ($vmid, $deviceid) = @_;
4619 mon_cmd($vmid, "netdev_del
", id => $deviceid);
4622 sub qemu_usb_hotplug {
4623 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4627 # remove the old one first
4628 vm_deviceunplug($vmid, $conf, $deviceid);
4630 # check if xhci controller is necessary and available
4631 my $devicelist = vm_devices_list($vmid);
4633 if (!$devicelist->{xhci}) {
4634 my $pciaddr = print_pci_addr("xhci
", undef, $arch, $machine_type);
4635 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_qemu_xhci_controller($pciaddr));
4638 # print_usbdevice_full expects the parsed device
4639 my $d = parse_usb_device($device->{host});
4640 $d->{usb3} = $device->{usb3};
4643 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4646 sub qemu_cpu_hotplug {
4647 my ($vmid, $conf, $vcpus) = @_;
4649 my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4652 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4653 $sockets = $conf->{sockets} if $conf->{sockets};
4654 my $cores = $conf->{cores} || 1;
4655 my $maxcpus = $sockets * $cores;
4657 $vcpus = $maxcpus if !$vcpus;
4659 die "you can
't add more vcpus than maxcpus\n"
4660 if $vcpus > $maxcpus;
4662 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4664 if ($vcpus < $currentvcpus) {
4666 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4668 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4669 qemu_devicedel($vmid, "cpu$i");
4671 my $currentrunningvcpus = undef;
4673 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4674 last if scalar(@{$currentrunningvcpus}) == $i-1;
4675 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4679 #update conf after each succesfull cpu unplug
4680 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4681 PVE::QemuConfig->write_config($vmid, $conf);
4684 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4690 my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4691 die "vcpus in running vm does not match its configuration\n"
4692 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4694 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4696 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4697 my $cpustr = print_cpu_device($conf, $i);
4698 qemu_deviceadd($vmid, $cpustr);
4701 my $currentrunningvcpus = undef;
4703 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4704 last if scalar(@{$currentrunningvcpus}) == $i;
4705 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4709 #update conf after each succesfull cpu hotplug
4710 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4711 PVE::QemuConfig->write_config($vmid, $conf);
4715 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4716 mon_cmd($vmid, "cpu-add", id => int($i));
4721 sub qemu_block_set_io_throttle {
4722 my ($vmid, $deviceid,
4723 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4724 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4725 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4726 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4728 return if !check_running($vmid) ;
4730 mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4732 bps_rd => int($bps_rd),
4733 bps_wr => int($bps_wr),
4735 iops_rd => int($iops_rd),
4736 iops_wr => int($iops_wr),
4737 bps_max => int($bps_max),
4738 bps_rd_max => int($bps_rd_max),
4739 bps_wr_max => int($bps_wr_max),
4740 iops_max => int($iops_max),
4741 iops_rd_max => int($iops_rd_max),
4742 iops_wr_max => int($iops_wr_max),
4743 bps_max_length => int($bps_max_length),
4744 bps_rd_max_length => int($bps_rd_max_length),
4745 bps_wr_max_length => int($bps_wr_max_length),
4746 iops_max_length => int($iops_max_length),
4747 iops_rd_max_length => int($iops_rd_max_length),
4748 iops_wr_max_length => int($iops_wr_max_length),
4753 sub qemu_block_resize {
4754 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4756 my $running = check_running($vmid);
4758 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4760 return if !$running;
4762 my $padding = (1024 - $size % 1024) % 1024;
4763 $size = $size + $padding;
4768 device => $deviceid,
4774 sub qemu_volume_snapshot {
4775 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4777 my $running = check_running($vmid);
4779 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4780 mon_cmd($vmid, 'blockdev-snapshot-internal-sync
', device => $deviceid, name => $snap);
4782 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4786 sub qemu_volume_snapshot_delete {
4787 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4789 my $running = check_running($vmid);
4794 my $conf = PVE::QemuConfig->load_config($vmid);
4795 PVE::QemuConfig->foreach_volume($conf, sub {
4796 my ($ds, $drive) = @_;
4797 $running = 1 if $drive->{file} eq $volid;
4801 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4802 mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync
', device => $deviceid, name => $snap);
4804 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4808 sub set_migration_caps {
4809 my ($vmid, $savevm) = @_;
4811 my $qemu_support = eval { mon_cmd($vmid, "query-proxmox-support") };
4813 my $bitmap_prop = $savevm ? 'pbs-dirty-bitmap-savevm
' : 'pbs-dirty-bitmap-migration
';
4814 my $dirty_bitmaps = $qemu_support->{$bitmap_prop} ? 1 : 0;
4819 "auto-converge" => 1,
4821 "x-rdma-pin-all" => 0,
4824 "dirty-bitmaps" => $dirty_bitmaps,
4827 my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4829 for my $supported_capability (@$supported_capabilities) {
4831 capability => $supported_capability->{capability},
4832 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4836 mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4840 my ($conf, $func, @param) = @_;
4844 my $test_volid = sub {
4845 my ($key, $drive, $snapname) = @_;
4847 my $volid = $drive->{file};
4850 $volhash->{$volid}->{cdrom} //= 1;
4851 $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4853 my $replicate = $drive->{replicate} // 1;
4854 $volhash->{$volid}->{replicate} //= 0;
4855 $volhash->{$volid}->{replicate} = 1 if $replicate;
4857 $volhash->{$volid}->{shared} //= 0;
4858 $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4860 $volhash->{$volid}->{referenced_in_config} //= 0;
4861 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
4863 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4864 if defined($snapname);
4866 my $size = $drive->{size};
4867 $volhash->{$volid}->{size} //= $size if $size;
4869 $volhash->{$volid}->{is_vmstate} //= 0;
4870 $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate
';
4872 $volhash->{$volid}->{is_tpmstate} //= 0;
4873 $volhash->{$volid}->{is_tpmstate} = 1 if $key eq 'tpmstate0
';
4875 $volhash->{$volid}->{is_unused} //= 0;
4876 $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4878 $volhash->{$volid}->{drivename} = $key if is_valid_drivename($key);
4881 my $include_opts = {
4882 extra_keys => ['vmstate
'],
4883 include_unused => 1,
4886 PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4887 foreach my $snapname (keys %{$conf->{snapshots}}) {
4888 my $snap = $conf->{snapshots}->{$snapname};
4889 PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4892 foreach my $volid (keys %$volhash) {
4893 &$func($volid, $volhash->{$volid}, @param);
4897 my $fast_plug_option = {
4905 'vmstatestorage
' => 1,
4910 for my $opt (keys %$confdesc_cloudinit) {
4911 $fast_plug_option->{$opt} = 1;
4914 # hotplug changes in [PENDING]
4915 # $selection hash can be used to only apply specified options, for
4916 # example: { cores => 1 } (only apply changed 'cores
')
4917 # $errors ref is used to return error messages
4918 sub vmconfig_hotplug_pending {
4919 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4921 my $defaults = load_defaults();
4922 my $arch = get_vm_arch($conf);
4923 my $machine_type = get_vm_machine($conf, undef, $arch);
4925 # commit values which do not have any impact on running VM first
4926 # Note: those option cannot raise errors, we we do not care about
4927 # $selection and always apply them.
4929 my $add_error = sub {
4930 my ($opt, $msg) = @_;
4931 $errors->{$opt} = "hotplug problem - $msg";
4934 my $cloudinit_pending_properties = PVE::QemuServer::cloudinit_pending_properties();
4936 my $cloudinit_record_changed = sub {
4937 my ($conf, $opt, $old, $new) = @_;
4938 return if !$cloudinit_pending_properties->{$opt};
4940 my $ci = ($conf->{cloudinit} //= {});
4942 my $recorded = $ci->{$opt};
4943 my %added = map { $_ => 1 } PVE::Tools::split_list(delete($ci->{added}) // '');
4945 if (defined($new)) {
4946 if (defined($old)) {
4947 # an existing value is being modified
4948 if (defined($recorded)) {
4949 # the value was already not in sync
4950 if ($new eq $recorded) {
4951 # a value is being reverted to the cloud-init state:
4953 delete $added{$opt};
4955 # the value was changed multiple times, do nothing
4957 } elsif ($added{$opt}) {
4958 # the value had been marked as added and is being changed, do nothing
4960 # the value is new, record it:
4964 # a new value is being added
4965 if (defined($recorded)) {
4966 # it was already not in sync
4967 if ($new eq $recorded) {
4968 # a value is being reverted to the cloud-init state:
4970 delete $added{$opt};
4972 # the value had temporarily been removed, do nothing
4974 } elsif ($added{$opt}) {
4975 # the value had been marked as added already, do nothing
4977 # the value is new, add it
4981 } elsif (!defined($old)) {
4982 # a non-existent value is being removed? ignore...
4984 # a value is being deleted
4985 if (defined($recorded)) {
4986 # a value was already recorded, just keep it
4987 } elsif ($added{$opt}) {
4988 # the value was marked as added, remove it
4989 delete $added{$opt};
4991 # a previously unrecorded value is being removed, record the old value:
4996 my $added = join(',', sort keys %added);
4997 $ci->{added} = $added if length($added);
5001 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5002 if ($fast_plug_option->{$opt}) {
5003 my $new = delete $conf->{pending}->{$opt};
5004 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $new);
5005 $conf->{$opt} = $new;
5011 PVE::QemuConfig->write_config($vmid, $conf);
5014 my $ostype = $conf->{ostype};
5015 my $version = extract_version($machine_type, get_running_qemu_version($vmid));
5016 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
5017 my $usb_hotplug = $hotplug_features->{usb}
5018 && min_version($version, 7, 1)
5019 && defined($ostype) && ($ostype eq 'l26
' || windows_version($ostype) > 7);
5021 my $cgroup = PVE::QemuServer::CGroup->new($vmid);
5022 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
5024 foreach my $opt (sort keys %$pending_delete_hash) {
5025 next if $selection && !$selection->{$opt};
5026 my $force = $pending_delete_hash->{$opt}->{force};
5028 if ($opt eq 'hotplug
') {
5029 die "skip\n" if ($conf->{hotplug} =~ /memory/);
5030 } elsif ($opt eq 'tablet
') {
5031 die "skip\n" if !$hotplug_features->{usb};
5032 if ($defaults->{tablet}) {
5033 vm_deviceplug($storecfg, $conf, $vmid, 'tablet
', $arch, $machine_type);
5034 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard
', $arch, $machine_type)
5035 if $arch eq 'aarch64
';
5037 vm_deviceunplug($vmid, $conf, 'tablet
');
5038 vm_deviceunplug($vmid, $conf, 'keyboard
') if $arch eq 'aarch64
';
5040 } elsif ($opt =~ m/^usb(\d+)$/) {
5042 die "skip\n" if !$usb_hotplug;
5043 vm_deviceunplug($vmid, $conf, "usbredirdev$index"); # if it's a spice port
5044 vm_deviceunplug
($vmid, $conf, $opt);
5045 } elsif ($opt eq 'vcpus') {
5046 die "skip\n" if !$hotplug_features->{cpu
};
5047 qemu_cpu_hotplug
($vmid, $conf, undef);
5048 } elsif ($opt eq 'balloon') {
5049 # enable balloon device is not hotpluggable
5050 die "skip\n" if defined($conf->{balloon
}) && $conf->{balloon
} == 0;
5051 # here we reset the ballooning value to memory
5052 my $balloon = $conf->{memory
} || $defaults->{memory
};
5053 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5054 } elsif ($fast_plug_option->{$opt}) {
5056 } elsif ($opt =~ m/^net(\d+)$/) {
5057 die "skip\n" if !$hotplug_features->{network
};
5058 vm_deviceunplug
($vmid, $conf, $opt);
5059 } elsif (is_valid_drivename
($opt)) {
5060 die "skip\n" if !$hotplug_features->{disk
} || $opt =~ m/(ide|sata)(\d+)/;
5061 vm_deviceunplug
($vmid, $conf, $opt);
5062 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5063 } elsif ($opt =~ m/^memory$/) {
5064 die "skip\n" if !$hotplug_features->{memory
};
5065 PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults);
5066 } elsif ($opt eq 'cpuunits') {
5067 $cgroup->change_cpu_shares(undef);
5068 } elsif ($opt eq 'cpulimit') {
5069 $cgroup->change_cpu_quota(undef, undef); # reset, cgroup module can better decide values
5075 &$add_error($opt, $err) if $err ne "skip\n";
5077 my $old = delete $conf->{$opt};
5078 $cloudinit_record_changed->($conf, $opt, $old, undef);
5079 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5084 foreach my $opt (keys %{$conf->{pending
}}) {
5085 next if $selection && !$selection->{$opt};
5086 my $value = $conf->{pending
}->{$opt};
5088 if ($opt eq 'hotplug') {
5089 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug
} =~ /memory/);
5090 } elsif ($opt eq 'tablet') {
5091 die "skip\n" if !$hotplug_features->{usb
};
5093 vm_deviceplug
($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5094 vm_deviceplug
($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5095 if $arch eq 'aarch64';
5096 } elsif ($value == 0) {
5097 vm_deviceunplug
($vmid, $conf, 'tablet');
5098 vm_deviceunplug
($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5100 } elsif ($opt =~ m/^usb(\d+)$/) {
5102 die "skip\n" if !$usb_hotplug;
5103 my $d = eval { parse_property_string
($usbdesc->{format
}, $value) };
5105 if ($d->{host
} eq 'spice') {
5106 $id = "usbredirdev$index";
5108 qemu_usb_hotplug
($storecfg, $conf, $vmid, $id, $d, $arch, $machine_type);
5109 } elsif ($opt eq 'vcpus') {
5110 die "skip\n" if !$hotplug_features->{cpu
};
5111 qemu_cpu_hotplug
($vmid, $conf, $value);
5112 } elsif ($opt eq 'balloon') {
5113 # enable/disable balloning device is not hotpluggable
5114 my $old_balloon_enabled = !!(!defined($conf->{balloon
}) || $conf->{balloon
});
5115 my $new_balloon_enabled = !!(!defined($conf->{pending
}->{balloon
}) || $conf->{pending
}->{balloon
});
5116 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5118 # allow manual ballooning if shares is set to zero
5119 if ((defined($conf->{shares
}) && ($conf->{shares
} == 0))) {
5120 my $balloon = $conf->{pending
}->{balloon
} || $conf->{memory
} || $defaults->{memory
};
5121 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5123 } elsif ($opt =~ m/^net(\d+)$/) {
5124 # some changes can be done without hotplug
5125 vmconfig_update_net
($storecfg, $conf, $hotplug_features->{network
},
5126 $vmid, $opt, $value, $arch, $machine_type);
5127 } elsif (is_valid_drivename
($opt)) {
5128 die "skip\n" if $opt eq 'efidisk0' || $opt eq 'tpmstate0';
5129 # some changes can be done without hotplug
5130 my $drive = parse_drive
($opt, $value);
5131 if (drive_is_cloudinit
($drive)) {
5132 $cloudinit_opt = [$opt, $drive];
5133 # apply all the other changes first, then generate the cloudinit disk
5136 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5137 $vmid, $opt, $value, $arch, $machine_type);
5138 } elsif ($opt =~ m/^memory$/) { #dimms
5139 die "skip\n" if !$hotplug_features->{memory
};
5140 $value = PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults, $value);
5141 } elsif ($opt eq 'cpuunits') {
5142 my $new_cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{pending
}->{$opt}); #clamp
5143 $cgroup->change_cpu_shares($new_cpuunits);
5144 } elsif ($opt eq 'cpulimit') {
5145 my $cpulimit = $conf->{pending
}->{$opt} == 0 ?
-1 : int($conf->{pending
}->{$opt} * 100000);
5146 $cgroup->change_cpu_quota($cpulimit, 100000);
5147 } elsif ($opt eq 'agent') {
5148 vmconfig_update_agent
($conf, $opt, $value);
5150 die "skip\n"; # skip non-hot-pluggable options
5154 &$add_error($opt, $err) if $err ne "skip\n";
5156 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $value);
5157 $conf->{$opt} = $value;
5158 delete $conf->{pending
}->{$opt};
5162 if (defined($cloudinit_opt)) {
5163 my ($opt, $drive) = @$cloudinit_opt;
5164 my $value = $conf->{pending
}->{$opt};
5166 my $temp = {%$conf, $opt => $value};
5167 PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($temp, $vmid);
5168 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5169 $vmid, $opt, $value, $arch, $machine_type);
5172 &$add_error($opt, $err) if $err ne "skip\n";
5174 $conf->{$opt} = $value;
5175 delete $conf->{pending
}->{$opt};
5179 # unplug xhci controller if no usb device is left
5182 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
5183 next if !defined($conf->{"usb$i"});
5188 vm_deviceunplug
($vmid, $conf, 'xhci');
5192 PVE
::QemuConfig-
>write_config($vmid, $conf);
5194 if ($hotplug_features->{cloudinit
} && PVE
::QemuServer
::Cloudinit
::has_changes
($conf)) {
5195 PVE
::QemuServer
::vmconfig_update_cloudinit_drive
($storecfg, $conf, $vmid);
5199 sub try_deallocate_drive
{
5200 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5202 if (($force || $key =~ /^unused/) && !drive_is_cdrom
($drive, 1)) {
5203 my $volid = $drive->{file
};
5204 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
5205 my $sid = PVE
::Storage
::parse_volume_id
($volid);
5206 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5208 # check if the disk is really unused
5209 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5210 if PVE
::QemuServer
::Drive
::is_volume_in_use
($storecfg, $conf, $key, $volid);
5211 PVE
::Storage
::vdisk_free
($storecfg, $volid);
5214 # If vm is not owner of this disk remove from config
5222 sub vmconfig_delete_or_detach_drive
{
5223 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5225 my $drive = parse_drive
($opt, $conf->{$opt});
5227 my $rpcenv = PVE
::RPCEnvironment
::get
();
5228 my $authuser = $rpcenv->get_user();
5231 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5232 try_deallocate_drive
($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5234 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $drive);
5240 sub vmconfig_apply_pending
{
5241 my ($vmid, $conf, $storecfg, $errors, $skip_cloud_init) = @_;
5243 return if !scalar(keys %{$conf->{pending
}});
5245 my $add_apply_error = sub {
5246 my ($opt, $msg) = @_;
5247 my $err_msg = "unable to apply pending change $opt : $msg";
5248 $errors->{$opt} = $err_msg;
5254 my $pending_delete_hash = PVE
::QemuConfig-
>parse_pending_delete($conf->{pending
}->{delete});
5255 foreach my $opt (sort keys %$pending_delete_hash) {
5256 my $force = $pending_delete_hash->{$opt}->{force
};
5258 if ($opt =~ m/^unused/) {
5259 die "internal error";
5260 } elsif (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5261 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5265 $add_apply_error->($opt, $err);
5267 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5268 delete $conf->{$opt};
5272 PVE
::QemuConfig-
>cleanup_pending($conf);
5274 my $generate_cloudinit = $skip_cloud_init ?
0 : undef;
5276 foreach my $opt (keys %{$conf->{pending
}}) { # add/change
5277 next if $opt eq 'delete'; # just to be sure
5279 if (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5280 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, parse_drive
($opt, $conf->{$opt}))
5284 $add_apply_error->($opt, $err);
5287 if (is_valid_drivename
($opt)) {
5288 my $drive = parse_drive
($opt, $conf->{pending
}->{$opt});
5289 $generate_cloudinit //= 1 if drive_is_cloudinit
($drive);
5292 $conf->{$opt} = delete $conf->{pending
}->{$opt};
5296 # write all changes at once to avoid unnecessary i/o
5297 PVE
::QemuConfig-
>write_config($vmid, $conf);
5298 if ($generate_cloudinit) {
5299 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5300 # After successful generation and if there were changes to be applied, update the
5301 # config to drop the {cloudinit} entry.
5302 PVE
::QemuConfig-
>write_config($vmid, $conf);
5307 sub vmconfig_update_net
{
5308 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5310 my $newnet = parse_net
($value);
5312 if ($conf->{$opt}) {
5313 my $oldnet = parse_net
($conf->{$opt});
5315 if (safe_string_ne
($oldnet->{model
}, $newnet->{model
}) ||
5316 safe_string_ne
($oldnet->{macaddr
}, $newnet->{macaddr
}) ||
5317 safe_num_ne
($oldnet->{queues
}, $newnet->{queues
}) ||
5318 !($newnet->{bridge
} && $oldnet->{bridge
})) { # bridge/nat mode change
5320 # for non online change, we try to hot-unplug
5321 die "skip\n" if !$hotplug;
5322 vm_deviceunplug
($vmid, $conf, $opt);
5325 die "internal error" if $opt !~ m/net(\d+)/;
5326 my $iface = "tap${vmid}i$1";
5328 if (safe_string_ne
($oldnet->{bridge
}, $newnet->{bridge
}) ||
5329 safe_num_ne
($oldnet->{tag
}, $newnet->{tag
}) ||
5330 safe_string_ne
($oldnet->{trunks
}, $newnet->{trunks
}) ||
5331 safe_num_ne
($oldnet->{firewall
}, $newnet->{firewall
})) {
5332 PVE
::Network
::tap_unplug
($iface);
5335 PVE
::Network
::SDN
::Zones
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5337 PVE
::Network
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5339 } elsif (safe_num_ne
($oldnet->{rate
}, $newnet->{rate
})) {
5340 # Rate can be applied on its own but any change above needs to
5341 # include the rate in tap_plug since OVS resets everything.
5342 PVE
::Network
::tap_rate_limit
($iface, $newnet->{rate
});
5345 if (safe_string_ne
($oldnet->{link_down
}, $newnet->{link_down
})) {
5346 qemu_set_link_status
($vmid, $opt, !$newnet->{link_down
});
5354 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5360 sub vmconfig_update_agent
{
5361 my ($conf, $opt, $value) = @_;
5363 die "skip\n" if !$conf->{$opt};
5365 my $hotplug_options = { fstrim_cloned_disks
=> 1 };
5367 my $old_agent = parse_guest_agent
($conf);
5368 my $agent = parse_guest_agent
({$opt => $value});
5370 for my $option (keys %$agent) { # added/changed options
5371 next if defined($hotplug_options->{$option});
5372 die "skip\n" if safe_string_ne
($agent->{$option}, $old_agent->{$option});
5375 for my $option (keys %$old_agent) { # removed options
5376 next if defined($hotplug_options->{$option});
5377 die "skip\n" if safe_string_ne
($old_agent->{$option}, $agent->{$option});
5380 return; # either no actual change (e.g., format string reordered) or just hotpluggable changes
5383 sub vmconfig_update_disk
{
5384 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5386 my $drive = parse_drive
($opt, $value);
5388 if ($conf->{$opt} && (my $old_drive = parse_drive
($opt, $conf->{$opt}))) {
5389 my $media = $drive->{media
} || 'disk';
5390 my $oldmedia = $old_drive->{media
} || 'disk';
5391 die "unable to change media type\n" if $media ne $oldmedia;
5393 if (!drive_is_cdrom
($old_drive)) {
5395 if ($drive->{file
} ne $old_drive->{file
}) {
5397 die "skip\n" if !$hotplug;
5399 # unplug and register as unused
5400 vm_deviceunplug
($vmid, $conf, $opt);
5401 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive)
5404 # update existing disk
5406 # skip non hotpluggable value
5407 if (safe_string_ne
($drive->{aio
}, $old_drive->{aio
}) ||
5408 safe_string_ne
($drive->{discard
}, $old_drive->{discard
}) ||
5409 safe_string_ne
($drive->{iothread
}, $old_drive->{iothread
}) ||
5410 safe_string_ne
($drive->{queues
}, $old_drive->{queues
}) ||
5411 safe_string_ne
($drive->{cache
}, $old_drive->{cache
}) ||
5412 safe_string_ne
($drive->{ssd
}, $old_drive->{ssd
}) ||
5413 safe_string_ne
($drive->{ro
}, $old_drive->{ro
})) {
5418 if (safe_num_ne
($drive->{mbps
}, $old_drive->{mbps
}) ||
5419 safe_num_ne
($drive->{mbps_rd
}, $old_drive->{mbps_rd
}) ||
5420 safe_num_ne
($drive->{mbps_wr
}, $old_drive->{mbps_wr
}) ||
5421 safe_num_ne
($drive->{iops
}, $old_drive->{iops
}) ||
5422 safe_num_ne
($drive->{iops_rd
}, $old_drive->{iops_rd
}) ||
5423 safe_num_ne
($drive->{iops_wr
}, $old_drive->{iops_wr
}) ||
5424 safe_num_ne
($drive->{mbps_max
}, $old_drive->{mbps_max
}) ||
5425 safe_num_ne
($drive->{mbps_rd_max
}, $old_drive->{mbps_rd_max
}) ||
5426 safe_num_ne
($drive->{mbps_wr_max
}, $old_drive->{mbps_wr_max
}) ||
5427 safe_num_ne
($drive->{iops_max
}, $old_drive->{iops_max
}) ||
5428 safe_num_ne
($drive->{iops_rd_max
}, $old_drive->{iops_rd_max
}) ||
5429 safe_num_ne
($drive->{iops_wr_max
}, $old_drive->{iops_wr_max
}) ||
5430 safe_num_ne
($drive->{bps_max_length
}, $old_drive->{bps_max_length
}) ||
5431 safe_num_ne
($drive->{bps_rd_max_length
}, $old_drive->{bps_rd_max_length
}) ||
5432 safe_num_ne
($drive->{bps_wr_max_length
}, $old_drive->{bps_wr_max_length
}) ||
5433 safe_num_ne
($drive->{iops_max_length
}, $old_drive->{iops_max_length
}) ||
5434 safe_num_ne
($drive->{iops_rd_max_length
}, $old_drive->{iops_rd_max_length
}) ||
5435 safe_num_ne
($drive->{iops_wr_max_length
}, $old_drive->{iops_wr_max_length
})) {
5437 qemu_block_set_io_throttle
(
5439 ($drive->{mbps
} || 0)*1024*1024,
5440 ($drive->{mbps_rd
} || 0)*1024*1024,
5441 ($drive->{mbps_wr
} || 0)*1024*1024,
5442 $drive->{iops
} || 0,
5443 $drive->{iops_rd
} || 0,
5444 $drive->{iops_wr
} || 0,
5445 ($drive->{mbps_max
} || 0)*1024*1024,
5446 ($drive->{mbps_rd_max
} || 0)*1024*1024,
5447 ($drive->{mbps_wr_max
} || 0)*1024*1024,
5448 $drive->{iops_max
} || 0,
5449 $drive->{iops_rd_max
} || 0,
5450 $drive->{iops_wr_max
} || 0,
5451 $drive->{bps_max_length
} || 1,
5452 $drive->{bps_rd_max_length
} || 1,
5453 $drive->{bps_wr_max_length
} || 1,
5454 $drive->{iops_max_length
} || 1,
5455 $drive->{iops_rd_max_length
} || 1,
5456 $drive->{iops_wr_max_length
} || 1,
5466 if ($drive->{file
} eq 'none') {
5467 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5468 if (drive_is_cloudinit
($old_drive)) {
5469 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive);
5472 my $path = get_iso_path
($storecfg, $vmid, $drive->{file
});
5474 # force eject if locked
5475 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5478 mon_cmd
($vmid, "blockdev-change-medium",
5479 id
=> "$opt", filename
=> "$path");
5487 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5489 PVE
::Storage
::activate_volumes
($storecfg, [$drive->{file
}]) if $drive->{file
} !~ m
|^/dev/.+|;
5490 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5493 sub vmconfig_update_cloudinit_drive
{
5494 my ($storecfg, $conf, $vmid) = @_;
5496 my $cloudinit_ds = undef;
5497 my $cloudinit_drive = undef;
5499 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5500 my ($ds, $drive) = @_;
5501 if (PVE
::QemuServer
::drive_is_cloudinit
($drive)) {
5502 $cloudinit_ds = $ds;
5503 $cloudinit_drive = $drive;
5507 return if !$cloudinit_drive;
5509 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5510 PVE
::QemuConfig-
>write_config($vmid, $conf);
5513 my $running = PVE
::QemuServer
::check_running
($vmid);
5516 my $path = PVE
::Storage
::path
($storecfg, $cloudinit_drive->{file
});
5518 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$cloudinit_ds");
5519 mon_cmd
($vmid, "blockdev-change-medium", id
=> "$cloudinit_ds", filename
=> "$path");
5524 # called in locked context by incoming migration
5525 sub vm_migrate_get_nbd_disks
{
5526 my ($storecfg, $conf, $replicated_volumes) = @_;
5528 my $local_volumes = {};
5529 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5530 my ($ds, $drive) = @_;
5532 return if drive_is_cdrom
($drive);
5533 return if $ds eq 'tpmstate0';
5535 my $volid = $drive->{file
};
5539 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
5541 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5542 return if $scfg->{shared
};
5544 # replicated disks re-use existing state via bitmap
5545 my $use_existing = $replicated_volumes->{$volid} ?
1 : 0;
5546 $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
5548 return $local_volumes;
5551 # called in locked context by incoming migration
5552 sub vm_migrate_alloc_nbd_disks
{
5553 my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
5556 foreach my $opt (sort keys %$source_volumes) {
5557 my ($volid, $storeid, $volname, $drive, $use_existing, $format) = @{$source_volumes->{$opt}};
5559 if ($use_existing) {
5560 $nbd->{$opt}->{drivestr
} = print_drive
($drive);
5561 $nbd->{$opt}->{volid
} = $volid;
5562 $nbd->{$opt}->{replicated
} = 1;
5566 # storage mapping + volname = regular migration
5567 # storage mapping + format = remote migration
5568 # order of precedence, filtered by whether storage supports it:
5569 # 1. explicit requested format
5570 # 2. format of current volume
5571 # 3. default format of storage
5572 if (!$storagemap->{identity
}) {
5573 $storeid = PVE
::JSONSchema
::map_id
($storagemap, $storeid);
5574 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
5575 if (!$format || !grep { $format eq $_ } @$validFormats) {
5577 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5578 my $fileFormat = qemu_img_format
($scfg, $volname);
5579 $format = $fileFormat
5580 if grep { $fileFormat eq $_ } @$validFormats;
5582 $format //= $defFormat;
5585 # can't happen for remote migration, so $volname is always defined
5586 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5587 $format = qemu_img_format
($scfg, $volname);
5590 my $size = $drive->{size
} / 1024;
5591 my $newvolid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $format, undef, $size);
5592 my $newdrive = $drive;
5593 $newdrive->{format
} = $format;
5594 $newdrive->{file
} = $newvolid;
5595 my $drivestr = print_drive
($newdrive);
5596 $nbd->{$opt}->{drivestr
} = $drivestr;
5597 $nbd->{$opt}->{volid
} = $newvolid;
5603 # see vm_start_nolock for parameters, additionally:
5605 # storagemap = parsed storage map for allocating NBD disks
5607 my ($storecfg, $vmid, $params, $migrate_opts) = @_;
5609 return PVE
::QemuConfig-
>lock_config($vmid, sub {
5610 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migrate_opts->{migratedfrom
});
5612 die "you can't start a vm if it's a template\n"
5613 if !$params->{skiptemplate
} && PVE
::QemuConfig-
>is_template($conf);
5615 my $has_suspended_lock = PVE
::QemuConfig-
>has_lock($conf, 'suspended');
5616 my $has_backup_lock = PVE
::QemuConfig-
>has_lock($conf, 'backup');
5618 my $running = check_running
($vmid, undef, $migrate_opts->{migratedfrom
});
5620 if ($has_backup_lock && $running) {
5621 # a backup is currently running, attempt to start the guest in the
5622 # existing QEMU instance
5623 return vm_resume
($vmid);
5626 PVE
::QemuConfig-
>check_lock($conf)
5627 if !($params->{skiplock
} || $has_suspended_lock);
5629 $params->{resume
} = $has_suspended_lock || defined($conf->{vmstate
});
5631 die "VM $vmid already running\n" if $running;
5633 if (my $storagemap = $migrate_opts->{storagemap
}) {
5634 my $replicated = $migrate_opts->{replicated_volumes
};
5635 my $disks = vm_migrate_get_nbd_disks
($storecfg, $conf, $replicated);
5636 $migrate_opts->{nbd
} = vm_migrate_alloc_nbd_disks
($storecfg, $vmid, $disks, $storagemap);
5638 foreach my $opt (keys %{$migrate_opts->{nbd
}}) {
5639 $conf->{$opt} = $migrate_opts->{nbd
}->{$opt}->{drivestr
};
5643 return vm_start_nolock
($storecfg, $vmid, $conf, $params, $migrate_opts);
5649 # statefile => 'tcp', 'unix' for migration or path/volid for RAM state
5650 # skiplock => 0/1, skip checking for config lock
5651 # skiptemplate => 0/1, skip checking whether VM is template
5652 # forcemachine => to force QEMU machine (rollback/migration)
5653 # forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
5654 # timeout => in seconds
5655 # paused => start VM in paused state (backup)
5656 # resume => resume from hibernation
5667 # nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
5668 # migratedfrom => source node
5669 # spice_ticket => used for spice migration, passed via tunnel/stdin
5670 # network => CIDR of migration network
5671 # type => secure/insecure - tunnel over encrypted connection or plain-text
5672 # nbd_proto_version => int, 0 for TCP, 1 for UNIX
5673 # replicated_volumes => which volids should be re-used with bitmaps for nbd migration
5674 # offline_volumes => new volids of offline migrated disks like tpmstate and cloudinit, not yet
5675 # contained in config
5676 sub vm_start_nolock
{
5677 my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
5679 my $statefile = $params->{statefile
};
5680 my $resume = $params->{resume
};
5682 my $migratedfrom = $migrate_opts->{migratedfrom
};
5683 my $migration_type = $migrate_opts->{type
};
5687 # clean up leftover reboot request files
5688 eval { clear_reboot_request
($vmid); };
5691 if (!$statefile && scalar(keys %{$conf->{pending
}})) {
5692 vmconfig_apply_pending
($vmid, $conf, $storecfg);
5693 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5696 # don't regenerate the ISO if the VM is started as part of a live migration
5697 # this way we can reuse the old ISO with the correct config
5698 if (!$migratedfrom) {
5699 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5700 # FIXME: apply_cloudinit_config updates $conf in this case, and it would only drop
5701 # $conf->{cloudinit}, so we could just not do this?
5702 # But we do it above, so for now let's be consistent.
5703 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5707 # override offline migrated volumes, conf is out of date still
5708 if (my $offline_volumes = $migrate_opts->{offline_volumes
}) {
5709 for my $key (sort keys $offline_volumes->%*) {
5710 my $parsed = parse_drive
($key, $conf->{$key});
5711 $parsed->{file
} = $offline_volumes->{$key};
5712 $conf->{$key} = print_drive
($parsed);
5716 my $defaults = load_defaults
();
5718 # set environment variable useful inside network script
5719 # for remote migration the config is available on the target node!
5720 if (!$migrate_opts->{remote_node
}) {
5721 $ENV{PVE_MIGRATED_FROM
} = $migratedfrom;
5724 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-start', 1);
5726 my $forcemachine = $params->{forcemachine
};
5727 my $forcecpu = $params->{forcecpu
};
5729 # enforce machine and CPU type on suspended vm to ensure HW compatibility
5730 $forcemachine = $conf->{runningmachine
};
5731 $forcecpu = $conf->{runningcpu
};
5732 print "Resuming suspended VM\n";
5735 my ($cmd, $vollist, $spice_port) = config_to_command
($storecfg, $vmid,
5736 $conf, $defaults, $forcemachine, $forcecpu, $params->{'pbs-backing'});
5739 my $get_migration_ip = sub {
5740 my ($nodename) = @_;
5742 return $migration_ip if defined($migration_ip);
5744 my $cidr = $migrate_opts->{network
};
5746 if (!defined($cidr)) {
5747 my $dc_conf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5748 $cidr = $dc_conf->{migration
}->{network
};
5751 if (defined($cidr)) {
5752 my $ips = PVE
::Network
::get_local_ip_from_cidr
($cidr);
5754 die "could not get IP: no address configured on local " .
5755 "node for network '$cidr'\n" if scalar(@$ips) == 0;
5757 die "could not get IP: multiple addresses configured on local " .
5758 "node for network '$cidr'\n" if scalar(@$ips) > 1;
5760 $migration_ip = @$ips[0];
5763 $migration_ip = PVE
::Cluster
::remote_node_ip
($nodename, 1)
5764 if !defined($migration_ip);
5766 return $migration_ip;
5770 if ($statefile eq 'tcp') {
5771 my $migrate = $res->{migrate
} = { proto
=> 'tcp' };
5772 $migrate->{addr
} = "localhost";
5773 my $datacenterconf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5774 my $nodename = nodename
();
5776 if (!defined($migration_type)) {
5777 if (defined($datacenterconf->{migration
}->{type
})) {
5778 $migration_type = $datacenterconf->{migration
}->{type
};
5780 $migration_type = 'secure';
5784 if ($migration_type eq 'insecure') {
5785 $migrate->{addr
} = $get_migration_ip->($nodename);
5786 $migrate->{addr
} = "[$migrate->{addr}]" if Net
::IP
::ip_is_ipv6
($migrate->{addr
});
5789 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5790 $migrate->{port
} = PVE
::Tools
::next_migrate_port
($pfamily);
5791 $migrate->{uri
} = "tcp:$migrate->{addr}:$migrate->{port}";
5792 push @$cmd, '-incoming', $migrate->{uri
};
5795 } elsif ($statefile eq 'unix') {
5796 # should be default for secure migrations as a ssh TCP forward
5797 # tunnel is not deterministic reliable ready and fails regurarly
5798 # to set up in time, so use UNIX socket forwards
5799 my $migrate = $res->{migrate
} = { proto
=> 'unix' };
5800 $migrate->{addr
} = "/run/qemu-server/$vmid.migrate";
5801 unlink $migrate->{addr
};
5803 $migrate->{uri
} = "unix:$migrate->{addr}";
5804 push @$cmd, '-incoming', $migrate->{uri
};
5807 } elsif (-e
$statefile) {
5808 push @$cmd, '-loadstate', $statefile;
5810 my $statepath = PVE
::Storage
::path
($storecfg, $statefile);
5811 push @$vollist, $statefile;
5812 push @$cmd, '-loadstate', $statepath;
5814 } elsif ($params->{paused
}) {
5818 my $start_timeout = $params->{timeout
} // config_aware_timeout
($conf, $resume);
5820 my $pci_devices = {}; # host pci devices
5821 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
5822 my $dev = $conf->{"hostpci$i"} or next;
5823 $pci_devices->{$i} = parse_hostpci
($dev);
5826 # do not reserve pciid for mediated devices, sysfs will error out for duplicate assignment
5827 my $real_pci_devices = [ grep { !(defined($_->{mdev
}) && scalar($_->{pciid
}->@*) == 1) } values $pci_devices->%* ];
5829 # map to a flat list of pci ids
5830 my $pci_id_list = [ map { $_->{id
} } map { $_->{pciid
}->@* } $real_pci_devices->@* ];
5832 # reserve all PCI IDs before actually doing anything with them
5833 PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, $start_timeout);
5837 for my $id (sort keys %$pci_devices) {
5838 my $d = $pci_devices->{$id};
5839 for my $dev ($d->{pciid
}->@*) {
5840 my $info = PVE
::QemuServer
::PCI
::prepare_pci_device
($vmid, $dev->{id
}, $id, $d->{mdev
});
5842 # nvidia grid needs the uuid of the mdev as qemu parameter
5843 if ($d->{mdev
} && !defined($uuid) && $info->{vendor
} eq '10de') {
5844 $uuid = PVE
::QemuServer
::PCI
::generate_mdev_uuid
($vmid, $id);
5848 push @$cmd, '-uuid', $uuid if defined($uuid);
5851 eval { cleanup_pci_devices
($vmid, $conf) };
5856 PVE
::Storage
::activate_volumes
($storecfg, $vollist);
5859 run_command
(['/bin/systemctl', 'stop', "$vmid.scope"], outfunc
=> sub{}, errfunc
=> sub{});
5861 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5862 # timeout should be more than enough here...
5863 PVE
::Systemd
::wait_for_unit_removed
("$vmid.scope", 20);
5865 my $cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{cpuunits
});
5868 timeout
=> $statefile ?
undef : $start_timeout,
5873 # when migrating, prefix QEMU output so other side can pick up any
5874 # errors that might occur and show the user
5875 if ($migratedfrom) {
5876 $run_params{quiet
} = 1;
5877 $run_params{logfunc
} = sub { print "QEMU: $_[0]\n" };
5880 my %systemd_properties = (
5881 Slice
=> 'qemu.slice',
5882 KillMode
=> 'process',
5884 TimeoutStopUSec
=> ULONG_MAX
, # infinity
5887 if (PVE
::CGroup
::cgroup_mode
() == 2) {
5888 $systemd_properties{CPUWeight
} = $cpuunits;
5890 $systemd_properties{CPUShares
} = $cpuunits;
5893 if (my $cpulimit = $conf->{cpulimit
}) {
5894 $systemd_properties{CPUQuota
} = int($cpulimit * 100);
5896 $systemd_properties{timeout
} = 10 if $statefile; # setting up the scope shoul be quick
5898 my $run_qemu = sub {
5899 PVE
::Tools
::run_fork
sub {
5900 PVE
::Systemd
::enter_systemd_scope
($vmid, "Proxmox VE VM $vmid", %systemd_properties);
5903 if (my $tpm = $conf->{tpmstate0
}) {
5904 # start the TPM emulator so QEMU can connect on start
5905 $tpmpid = start_swtpm
($storecfg, $vmid, $tpm, $migratedfrom);
5908 my $exitcode = run_command
($cmd, %run_params);
5911 warn "stopping swtpm instance (pid $tpmpid) due to QEMU startup error\n";
5912 kill 'TERM', $tpmpid;
5914 die "QEMU exited with code $exitcode\n";
5919 if ($conf->{hugepages
}) {
5922 my $hotplug_features =
5923 parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
5924 my $hugepages_topology =
5925 PVE
::QemuServer
::Memory
::hugepages_topology
($conf, $hotplug_features->{memory
});
5927 my $hugepages_host_topology = PVE
::QemuServer
::Memory
::hugepages_host_topology
();
5929 PVE
::QemuServer
::Memory
::hugepages_mount
();
5930 PVE
::QemuServer
::Memory
::hugepages_allocate
($hugepages_topology, $hugepages_host_topology);
5932 eval { $run_qemu->() };
5934 PVE
::QemuServer
::Memory
::hugepages_reset
($hugepages_host_topology)
5935 if !$conf->{keephugepages
};
5939 PVE
::QemuServer
::Memory
::hugepages_pre_deallocate
($hugepages_topology)
5940 if !$conf->{keephugepages
};
5942 eval { PVE
::QemuServer
::Memory
::hugepages_update_locked
($code); };
5945 eval { $run_qemu->() };
5949 # deactivate volumes if start fails
5950 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
5952 eval { cleanup_pci_devices
($vmid, $conf) };
5955 die "start failed: $err";
5958 # re-reserve all PCI IDs now that we can know the actual VM PID
5959 my $pid = PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
5960 eval { PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, undef, $pid) };
5963 if (defined($res->{migrate
})) {
5964 print "migration listens on $res->{migrate}->{uri}\n";
5965 } elsif ($statefile) {
5966 eval { mon_cmd
($vmid, "cont"); };
5970 #start nbd server for storage migration
5971 if (my $nbd = $migrate_opts->{nbd
}) {
5972 my $nbd_protocol_version = $migrate_opts->{nbd_proto_version
} // 0;
5974 my $migrate_storage_uri;
5975 # nbd_protocol_version > 0 for unix socket support
5976 if ($nbd_protocol_version > 0 && ($migration_type eq 'secure' || $migration_type eq 'websocket')) {
5977 my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
5978 mon_cmd
($vmid, "nbd-server-start", addr
=> { type
=> 'unix', data
=> { path
=> $socket_path } } );
5979 $migrate_storage_uri = "nbd:unix:$socket_path";
5980 $res->{migrate
}->{unix_sockets
} = [$socket_path];
5982 my $nodename = nodename
();
5983 my $localip = $get_migration_ip->($nodename);
5984 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5985 my $storage_migrate_port = PVE
::Tools
::next_migrate_port
($pfamily);
5987 mon_cmd
($vmid, "nbd-server-start", addr
=> {
5990 host
=> "${localip}",
5991 port
=> "${storage_migrate_port}",
5994 $localip = "[$localip]" if Net
::IP
::ip_is_ipv6
($localip);
5995 $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
5998 my $block_info = mon_cmd
($vmid, "query-block");
5999 $block_info = { map { $_->{device
} => $_ } $block_info->@* };
6001 foreach my $opt (sort keys %$nbd) {
6002 my $drivestr = $nbd->{$opt}->{drivestr
};
6003 my $volid = $nbd->{$opt}->{volid
};
6005 my $block_node = $block_info->{"drive-$opt"}->{inserted
}->{'node-name'};
6011 'node-name' => $block_node,
6012 writable
=> JSON
::true
,
6014 name
=> "drive-$opt", # NBD export name
6017 my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
6018 print "storage migration listens on $nbd_uri volume:$drivestr\n";
6019 print "re-using replicated volume: $opt - $volid\n"
6020 if $nbd->{$opt}->{replicated
};
6022 $res->{drives
}->{$opt} = $nbd->{$opt};
6023 $res->{drives
}->{$opt}->{nbd_uri
} = $nbd_uri;
6027 if ($migratedfrom) {
6029 set_migration_caps
($vmid);
6034 print "spice listens on port $spice_port\n";
6035 $res->{spice_port
} = $spice_port;
6036 if ($migrate_opts->{spice_ticket
}) {
6037 mon_cmd
($vmid, "set_password", protocol
=> 'spice', password
=>
6038 $migrate_opts->{spice_ticket
});
6039 mon_cmd
($vmid, "expire_password", protocol
=> 'spice', time => "+30");
6044 mon_cmd
($vmid, "balloon", value
=> $conf->{balloon
}*1024*1024)
6045 if !$statefile && $conf->{balloon
};
6047 foreach my $opt (keys %$conf) {
6048 next if $opt !~ m/^net\d+$/;
6049 my $nicconf = parse_net
($conf->{$opt});
6050 qemu_set_link_status
($vmid, $opt, 0) if $nicconf->{link_down
};
6052 add_nets_bridge_fdb
($conf, $vmid);
6055 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
6060 path
=> "machine/peripheral/balloon0",
6061 property
=> "guest-stats-polling-interval",
6065 log_warn
("could not set polling interval for ballooning - $@") if $@;
6069 print "Resumed VM, removing state\n";
6070 if (my $vmstate = $conf->{vmstate
}) {
6071 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6072 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6074 delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
6075 PVE
::QemuConfig-
>write_config($vmid, $conf);
6078 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'post-start');
6083 sub vm_commandline
{
6084 my ($storecfg, $vmid, $snapname) = @_;
6086 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6088 my ($forcemachine, $forcecpu);
6090 my $snapshot = $conf->{snapshots
}->{$snapname};
6091 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
6093 # check for machine or CPU overrides in snapshot
6094 $forcemachine = $snapshot->{runningmachine
};
6095 $forcecpu = $snapshot->{runningcpu
};
6097 $snapshot->{digest
} = $conf->{digest
}; # keep file digest for API
6102 my $defaults = load_defaults
();
6104 my $cmd = config_to_command
($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
6106 return PVE
::Tools
::cmd2string
($cmd);
6110 my ($vmid, $skiplock) = @_;
6112 PVE
::QemuConfig-
>lock_config($vmid, sub {
6114 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6116 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6118 mon_cmd
($vmid, "system_reset");
6122 sub get_vm_volumes
{
6126 foreach_volid
($conf, sub {
6127 my ($volid, $attr) = @_;
6129 return if $volid =~ m
|^/|;
6131 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
6134 push @$vollist, $volid;
6140 sub cleanup_pci_devices
{
6141 my ($vmid, $conf) = @_;
6143 foreach my $key (keys %$conf) {
6144 next if $key !~ m/^hostpci(\d+)$/;
6145 my $hostpciindex = $1;
6146 my $uuid = PVE
::SysFSTools
::generate_mdev_uuid
($vmid, $hostpciindex);
6147 my $d = parse_hostpci
($conf->{$key});
6149 # NOTE: avoid PVE::SysFSTools::pci_cleanup_mdev_device as it requires PCI ID and we
6150 # don't want to break ABI just for this two liner
6151 my $dev_sysfs_dir = "/sys/bus/mdev/devices/$uuid";
6152 PVE
::SysFSTools
::file_write
("$dev_sysfs_dir/remove", "1") if -e
$dev_sysfs_dir;
6155 PVE
::QemuServer
::PCI
::remove_pci_reservation
($vmid);
6158 sub vm_stop_cleanup
{
6159 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
6164 my $vollist = get_vm_volumes
($conf);
6165 PVE
::Storage
::deactivate_volumes
($storecfg, $vollist);
6167 if (my $tpmdrive = $conf->{tpmstate0
}) {
6168 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
6169 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
6171 PVE
::Storage
::unmap_volume
($storecfg, $tpm->{file
});
6176 foreach my $ext (qw(mon qmp pid vnc qga)) {
6177 unlink "/var/run/qemu-server/${vmid}.$ext";
6180 if ($conf->{ivshmem
}) {
6181 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
6182 # just delete it for now, VMs which have this already open do not
6183 # are affected, but new VMs will get a separated one. If this
6184 # becomes an issue we either add some sort of ref-counting or just
6185 # add a "don't delete on stop" flag to the ivshmem format.
6186 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name
} // $vmid);
6189 cleanup_pci_devices
($vmid, $conf);
6191 vmconfig_apply_pending
($vmid, $conf, $storecfg) if $apply_pending_changes;
6193 warn $@ if $@; # avoid errors - just warn
6196 # call only in locked context
6198 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
6200 my $pid = check_running
($vmid, $nocheck);
6205 $conf = PVE
::QemuConfig-
>load_config($vmid);
6206 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6207 if (!defined($timeout) && $shutdown && $conf->{startup
}) {
6208 my $opts = PVE
::JSONSchema
::pve_parse_startup_order
($conf->{startup
});
6209 $timeout = $opts->{down
} if $opts->{down
};
6211 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-stop');
6216 if (defined($conf) && get_qga_key
($conf, 'enabled')) {
6217 mon_cmd
($vmid, "guest-shutdown", timeout
=> $timeout);
6219 mon_cmd
($vmid, "system_powerdown");
6222 mon_cmd
($vmid, "quit");
6228 $timeout = 60 if !defined($timeout);
6231 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6236 if ($count >= $timeout) {
6238 warn "VM still running - terminating now with SIGTERM\n";
6241 die "VM quit/powerdown failed - got timeout\n";
6244 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6248 if (!check_running
($vmid, $nocheck)) {
6249 warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
6253 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
6256 die "VM quit/powerdown failed\n";
6264 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6269 if ($count >= $timeout) {
6270 warn "VM still running - terminating now with SIGKILL\n";
6275 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6278 # Note: use $nocheck to skip tests if VM configuration file exists.
6279 # We need that when migration VMs to other nodes (files already moved)
6280 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
6282 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
6284 $force = 1 if !defined($force) && !$shutdown;
6287 my $pid = check_running
($vmid, $nocheck, $migratedfrom);
6288 kill 15, $pid if $pid;
6289 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migratedfrom);
6290 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 0);
6294 PVE
::QemuConfig-
>lock_config($vmid, sub {
6295 _do_vm_stop
($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
6300 my ($vmid, $timeout) = @_;
6302 PVE
::QemuConfig-
>lock_config($vmid, sub {
6305 # only reboot if running, as qmeventd starts it again on a stop event
6306 return if !check_running
($vmid);
6308 create_reboot_request
($vmid);
6310 my $storecfg = PVE
::Storage
::config
();
6311 _do_vm_stop
($storecfg, $vmid, undef, undef, $timeout, 1);
6315 # avoid that the next normal shutdown will be confused for a reboot
6316 clear_reboot_request
($vmid);
6322 # note: if using the statestorage parameter, the caller has to check privileges
6324 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
6331 PVE
::QemuConfig-
>lock_config($vmid, sub {
6333 $conf = PVE
::QemuConfig-
>load_config($vmid);
6335 my $is_backing_up = PVE
::QemuConfig-
>has_lock($conf, 'backup');
6336 PVE
::QemuConfig-
>check_lock($conf)
6337 if !($skiplock || $is_backing_up);
6339 die "cannot suspend to disk during backup\n"
6340 if $is_backing_up && $includestate;
6342 if ($includestate) {
6343 $conf->{lock} = 'suspending';
6344 my $date = strftime
("%Y-%m-%d", localtime(time()));
6345 $storecfg = PVE
::Storage
::config
();
6346 if (!$statestorage) {
6347 $statestorage = find_vmstate_storage
($conf, $storecfg);
6348 # check permissions for the storage
6349 my $rpcenv = PVE
::RPCEnvironment
::get
();
6350 if ($rpcenv->{type
} ne 'cli') {
6351 my $authuser = $rpcenv->get_user();
6352 $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
6357 $vmstate = PVE
::QemuConfig-
>__snapshot_save_vmstate(
6358 $vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
6359 $path = PVE
::Storage
::path
($storecfg, $vmstate);
6360 PVE
::QemuConfig-
>write_config($vmid, $conf);
6362 mon_cmd
($vmid, "stop");
6366 if ($includestate) {
6368 PVE
::Storage
::activate_volumes
($storecfg, [$vmstate]);
6371 set_migration_caps
($vmid, 1);
6372 mon_cmd
($vmid, "savevm-start", statefile
=> $path);
6374 my $state = mon_cmd
($vmid, "query-savevm");
6375 if (!$state->{status
}) {
6376 die "savevm not active\n";
6377 } elsif ($state->{status
} eq 'active') {
6380 } elsif ($state->{status
} eq 'completed') {
6381 print "State saved, quitting\n";
6383 } elsif ($state->{status
} eq 'failed' && $state->{error
}) {
6384 die "query-savevm failed with error '$state->{error}'\n"
6386 die "query-savevm returned status '$state->{status}'\n";
6392 PVE
::QemuConfig-
>lock_config($vmid, sub {
6393 $conf = PVE
::QemuConfig-
>load_config($vmid);
6395 # cleanup, but leave suspending lock, to indicate something went wrong
6397 mon_cmd
($vmid, "savevm-end");
6398 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6399 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6400 delete $conf->@{qw(vmstate runningmachine runningcpu)};
6401 PVE
::QemuConfig-
>write_config($vmid, $conf);
6407 die "lock changed unexpectedly\n"
6408 if !PVE
::QemuConfig-
>has_lock($conf, 'suspending');
6410 mon_cmd
($vmid, "quit");
6411 $conf->{lock} = 'suspended';
6412 PVE
::QemuConfig-
>write_config($vmid, $conf);
6417 # $nocheck is set when called as part of a migration - in this context the
6418 # location of the config file (source or target node) is not deterministic,
6419 # since migration cannot wait for pmxcfs to process the rename
6421 my ($vmid, $skiplock, $nocheck) = @_;
6423 PVE
::QemuConfig-
>lock_config($vmid, sub {
6424 my $res = mon_cmd
($vmid, 'query-status');
6425 my $resume_cmd = 'cont';
6429 $conf = eval { PVE
::QemuConfig-
>load_config($vmid) }; # try on target node
6431 my $vmlist = PVE
::Cluster
::get_vmlist
();
6432 if (exists($vmlist->{ids
}->{$vmid})) {
6433 my $node = $vmlist->{ids
}->{$vmid}->{node
};
6434 $conf = eval { PVE
::QemuConfig-
>load_config($vmid, $node) }; # try on source node
6437 PVE
::Cluster
::cfs_update
(); # vmlist was wrong, invalidate cache
6438 $conf = PVE
::QemuConfig-
>load_config($vmid); # last try on target node again
6442 $conf = PVE
::QemuConfig-
>load_config($vmid);
6445 if ($res->{status
}) {
6446 return if $res->{status
} eq 'running'; # job done, go home
6447 $resume_cmd = 'system_wakeup' if $res->{status
} eq 'suspended';
6448 $reset = 1 if $res->{status
} eq 'shutdown';
6452 PVE
::QemuConfig-
>check_lock($conf)
6453 if !($skiplock || PVE
::QemuConfig-
>has_lock($conf, 'backup'));
6457 # required if a VM shuts down during a backup and we get a resume
6458 # request before the backup finishes for example
6459 mon_cmd
($vmid, "system_reset");
6462 add_nets_bridge_fdb
($conf, $vmid) if $resume_cmd eq 'cont';
6464 mon_cmd
($vmid, $resume_cmd);
6469 my ($vmid, $skiplock, $key) = @_;
6471 PVE
::QemuConfig-
>lock_config($vmid, sub {
6473 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6475 # there is no qmp command, so we use the human monitor command
6476 my $res = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "sendkey $key");
6477 die $res if $res ne '';
6481 # vzdump restore implementaion
6483 sub tar_archive_read_firstfile
{
6484 my $archive = shift;
6486 die "ERROR: file '$archive' does not exist\n" if ! -f
$archive;
6488 # try to detect archive type first
6489 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6490 die "unable to open file '$archive'\n";
6491 my $firstfile = <$fh>;
6495 die "ERROR: archive contaions no data\n" if !$firstfile;
6501 sub tar_restore_cleanup
{
6502 my ($storecfg, $statfile) = @_;
6504 print STDERR
"starting cleanup\n";
6506 if (my $fd = IO
::File-
>new($statfile, "r")) {
6507 while (defined(my $line = <$fd>)) {
6508 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6511 if ($volid =~ m
|^/|) {
6512 unlink $volid || die 'unlink failed\n';
6514 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6516 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6518 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6520 print STDERR
"unable to parse line in statfile - $line";
6527 sub restore_file_archive
{
6528 my ($archive, $vmid, $user, $opts) = @_;
6530 return restore_vma_archive
($archive, $vmid, $user, $opts)
6533 my $info = PVE
::Storage
::archive_info
($archive);
6534 my $format = $opts->{format
} // $info->{format
};
6535 my $comp = $info->{compression
};
6537 # try to detect archive format
6538 if ($format eq 'tar') {
6539 return restore_tar_archive
($archive, $vmid, $user, $opts);
6541 return restore_vma_archive
($archive, $vmid, $user, $opts, $comp);
6545 # hepler to remove disks that will not be used after restore
6546 my $restore_cleanup_oldconf = sub {
6547 my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
6549 my $kept_disks = {};
6551 PVE
::QemuConfig-
>foreach_volume($oldconf, sub {
6552 my ($ds, $drive) = @_;
6554 return if drive_is_cdrom
($drive, 1);
6556 my $volid = $drive->{file
};
6557 return if !$volid || $volid =~ m
|^/|;
6559 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
6560 return if !$path || !$owner || ($owner != $vmid);
6562 # Note: only delete disk we want to restore
6563 # other volumes will become unused
6564 if ($virtdev_hash->{$ds}) {
6565 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
6570 $kept_disks->{$volid} = 1;
6574 # after the restore we have no snapshots anymore
6575 for my $snapname (keys $oldconf->{snapshots
}->%*) {
6576 my $snap = $oldconf->{snapshots
}->{$snapname};
6577 if ($snap->{vmstate
}) {
6578 eval { PVE
::Storage
::vdisk_free
($storecfg, $snap->{vmstate
}); };
6584 for my $volid (keys $kept_disks->%*) {
6585 eval { PVE
::Storage
::volume_snapshot_delete
($storecfg, $volid, $snapname); };
6591 # Helper to parse vzdump backup device hints
6593 # $rpcenv: Environment, used to ckeck storage permissions
6594 # $user: User ID, to check storage permissions
6595 # $storecfg: Storage configuration
6596 # $fh: the file handle for reading the configuration
6597 # $devinfo: should contain device sizes for all backu-up'ed devices
6598 # $options: backup options (pool, default storage)
6600 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
6601 my $parse_backup_hints = sub {
6602 my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
6604 my $check_storage = sub { # assert if an image can be allocate
6605 my ($storeid, $scfg) = @_;
6606 die "Content type 'images' is not available on storage '$storeid'\n"
6607 if !$scfg->{content
}->{images
};
6608 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace'])
6609 if $user ne 'root@pam';
6612 my $virtdev_hash = {};
6613 while (defined(my $line = <$fh>)) {
6614 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6615 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6616 die "archive does not contain data for drive '$virtdev'\n"
6617 if !$devinfo->{$devname};
6619 if (defined($options->{storage
})) {
6620 $storeid = $options->{storage
} || 'local';
6621 } elsif (!$storeid) {
6624 $format = 'raw' if !$format;
6625 $devinfo->{$devname}->{devname
} = $devname;
6626 $devinfo->{$devname}->{virtdev
} = $virtdev;
6627 $devinfo->{$devname}->{format
} = $format;
6628 $devinfo->{$devname}->{storeid
} = $storeid;
6630 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6631 $check_storage->($storeid, $scfg); # permission and content type check
6633 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6634 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6636 my $drive = parse_drive
($virtdev, $2);
6638 if (drive_is_cloudinit
($drive)) {
6639 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
6640 $storeid = $options->{storage
} if defined ($options->{storage
});
6641 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6642 my $format = qemu_img_format
($scfg, $volname); # has 'raw' fallback
6644 $check_storage->($storeid, $scfg); # permission and content type check
6646 $virtdev_hash->{$virtdev} = {
6648 storeid
=> $storeid,
6649 size
=> PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
,
6656 return $virtdev_hash;
6659 # Helper to allocate and activate all volumes required for a restore
6661 # $storecfg: Storage configuration
6662 # $virtdev_hash: as returned by parse_backup_hints()
6664 # Returns: { $virtdev => $volid }
6665 my $restore_allocate_devices = sub {
6666 my ($storecfg, $virtdev_hash, $vmid) = @_;
6669 foreach my $virtdev (sort keys %$virtdev_hash) {
6670 my $d = $virtdev_hash->{$virtdev};
6671 my $alloc_size = int(($d->{size
} + 1024 - 1)/1024);
6672 my $storeid = $d->{storeid
};
6673 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6675 # test if requested format is supported
6676 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
6677 my $supported = grep { $_ eq $d->{format
} } @$validFormats;
6678 $d->{format
} = $defFormat if !$supported;
6681 if ($d->{is_cloudinit
}) {
6682 $name = "vm-$vmid-cloudinit";
6683 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6684 if ($scfg->{path
}) {
6685 $name .= ".$d->{format}";
6689 my $volid = PVE
::Storage
::vdisk_alloc
(
6690 $storecfg, $storeid, $vmid, $d->{format
}, $name, $alloc_size);
6692 print STDERR
"new volume ID is '$volid'\n";
6693 $d->{volid
} = $volid;
6695 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
6697 $map->{$virtdev} = $volid;
6703 sub restore_update_config_line
{
6704 my ($cookie, $map, $line, $unique) = @_;
6706 return '' if $line =~ m/^\#qmdump\#/;
6707 return '' if $line =~ m/^\#vzdump\#/;
6708 return '' if $line =~ m/^lock:/;
6709 return '' if $line =~ m/^unused\d+:/;
6710 return '' if $line =~ m/^parent:/;
6714 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
6715 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6716 # try to convert old 1.X settings
6717 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6718 foreach my $devconfig (PVE
::Tools
::split_list
($ethcfg)) {
6719 my ($model, $macaddr) = split(/\=/, $devconfig);
6720 $macaddr = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if !$macaddr || $unique;
6723 bridge
=> "vmbr$ind",
6724 macaddr
=> $macaddr,
6726 my $netstr = print_net
($net);
6728 $res .= "net$cookie->{netcount}: $netstr\n";
6729 $cookie->{netcount
}++;
6731 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6732 my ($id, $netstr) = ($1, $2);
6733 my $net = parse_net
($netstr);
6734 $net->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if $net->{macaddr
};
6735 $netstr = print_net
($net);
6736 $res .= "$id: $netstr\n";
6737 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk|tpmstate)\d+):\s*(\S+)\s*$/) {
6740 my $di = parse_drive
($virtdev, $value);
6741 if (defined($di->{backup
}) && !$di->{backup
}) {
6743 } elsif ($map->{$virtdev}) {
6744 delete $di->{format
}; # format can change on restore
6745 $di->{file
} = $map->{$virtdev};
6746 $value = print_drive
($di);
6747 $res .= "$virtdev: $value\n";
6751 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6753 if ($vmgenid ne '0') {
6754 # always generate a new vmgenid if there was a valid one setup
6755 $vmgenid = generate_uuid
();
6757 $res .= "vmgenid: $vmgenid\n";
6758 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6759 my ($uuid, $uuid_str);
6760 UUID
::generate
($uuid);
6761 UUID
::unparse
($uuid, $uuid_str);
6762 my $smbios1 = parse_smbios1
($2);
6763 $smbios1->{uuid
} = $uuid_str;
6764 $res .= $1.print_smbios1
($smbios1)."\n";
6772 my $restore_deactivate_volumes = sub {
6773 my ($storecfg, $virtdev_hash) = @_;
6776 for my $dev (values $virtdev_hash->%*) {
6777 push $vollist->@*, $dev->{volid
} if $dev->{volid
};
6780 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
6781 print STDERR
$@ if $@;
6784 my $restore_destroy_volumes = sub {
6785 my ($storecfg, $virtdev_hash) = @_;
6787 for my $dev (values $virtdev_hash->%*) {
6788 my $volid = $dev->{volid
} or next;
6790 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6791 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6793 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6797 my $restore_merge_config = sub {
6798 my ($filename, $backup_conf_raw, $override_conf) = @_;
6800 my $backup_conf = parse_vm_config
($filename, $backup_conf_raw);
6801 for my $key (keys $override_conf->%*) {
6802 $backup_conf->{$key} = $override_conf->{$key};
6805 return $backup_conf;
6809 my ($cfg, $vmid) = @_;
6811 my $info = PVE
::Storage
::vdisk_list
($cfg, undef, $vmid, undef, 'images');
6813 my $volid_hash = {};
6814 foreach my $storeid (keys %$info) {
6815 foreach my $item (@{$info->{$storeid}}) {
6816 next if !($item->{volid
} && $item->{size
});
6817 $item->{path
} = PVE
::Storage
::path
($cfg, $item->{volid
});
6818 $volid_hash->{$item->{volid
}} = $item;
6825 sub update_disk_config
{
6826 my ($vmid, $conf, $volid_hash) = @_;
6829 my $prefix = "VM $vmid";
6831 # used and unused disks
6832 my $referenced = {};
6834 # Note: it is allowed to define multiple storages with same path (alias), so
6835 # we need to check both 'volid' and real 'path' (two different volid can point
6836 # to the same path).
6838 my $referencedpath = {};
6841 PVE
::QemuConfig-
>foreach_volume($conf, sub {
6842 my ($opt, $drive) = @_;
6844 my $volid = $drive->{file
};
6846 my $volume = $volid_hash->{$volid};
6848 # mark volid as "in-use" for next step
6849 $referenced->{$volid} = 1;
6850 if ($volume && (my $path = $volume->{path
})) {
6851 $referencedpath->{$path} = 1;
6854 return if drive_is_cdrom
($drive);
6857 my ($updated, $msg) = PVE
::QemuServer
::Drive
::update_disksize
($drive, $volume->{size
});
6858 if (defined($updated)) {
6860 $conf->{$opt} = print_drive
($updated);
6861 print "$prefix ($opt): $msg\n";
6865 # remove 'unusedX' entry if volume is used
6866 PVE
::QemuConfig-
>foreach_unused_volume($conf, sub {
6867 my ($opt, $drive) = @_;
6869 my $volid = $drive->{file
};
6873 $path = $volid_hash->{$volid}->{path
} if $volid_hash->{$volid};
6874 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6875 print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
6877 delete $conf->{$opt};
6880 $referenced->{$volid} = 1;
6881 $referencedpath->{$path} = 1 if $path;
6884 foreach my $volid (sort keys %$volid_hash) {
6885 next if $volid =~ m/vm-$vmid-state-/;
6886 next if $referenced->{$volid};
6887 my $path = $volid_hash->{$volid}->{path
};
6888 next if !$path; # just to be sure
6889 next if $referencedpath->{$path};
6891 my $key = PVE
::QemuConfig-
>add_unused_volume($conf, $volid);
6892 print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
6893 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6900 my ($vmid, $nolock, $dryrun) = @_;
6902 my $cfg = PVE
::Storage
::config
();
6904 print "rescan volumes...\n";
6905 my $volid_hash = scan_volids
($cfg, $vmid);
6907 my $updatefn = sub {
6910 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6912 PVE
::QemuConfig-
>check_lock($conf);
6915 foreach my $volid (keys %$volid_hash) {
6916 my $info = $volid_hash->{$volid};
6917 $vm_volids->{$volid} = $info if $info->{vmid
} && $info->{vmid
} == $vmid;
6920 my $changes = update_disk_config
($vmid, $conf, $vm_volids);
6922 PVE
::QemuConfig-
>write_config($vmid, $conf) if $changes && !$dryrun;
6925 if (defined($vmid)) {
6929 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6932 my $vmlist = config_list
();
6933 foreach my $vmid (keys %$vmlist) {
6937 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6943 sub restore_proxmox_backup_archive
{
6944 my ($archive, $vmid, $user, $options) = @_;
6946 my $storecfg = PVE
::Storage
::config
();
6948 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($archive);
6949 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6951 my $fingerprint = $scfg->{fingerprint
};
6952 my $keyfile = PVE
::Storage
::PBSPlugin
::pbs_encryption_key_file_name
($storecfg, $storeid);
6954 my $repo = PVE
::PBSClient
::get_repository
($scfg);
6955 my $namespace = $scfg->{namespace
};
6957 # This is only used for `pbs-restore` and the QEMU PBS driver (live-restore)
6958 my $password = PVE
::Storage
::PBSPlugin
::pbs_get_password
($scfg, $storeid);
6959 local $ENV{PBS_PASSWORD
} = $password;
6960 local $ENV{PBS_FINGERPRINT
} = $fingerprint if defined($fingerprint);
6962 my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6963 PVE
::Storage
::parse_volname
($storecfg, $archive);
6965 die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
6967 die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
6969 my $tmpdir = "/var/tmp/vzdumptmp$$";
6973 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
6974 # disable interrupts (always do cleanups)
6978 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
6980 # Note: $oldconf is undef if VM does not exists
6981 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
6982 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
6983 my $new_conf_raw = '';
6985 my $rpcenv = PVE
::RPCEnvironment
::get
();
6986 my $devinfo = {}; # info about drives included in backup
6987 my $virtdev_hash = {}; # info about allocated drives
6995 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
6997 my $cfgfn = "$tmpdir/qemu-server.conf";
6998 my $firewall_config_fn = "$tmpdir/fw.conf";
6999 my $index_fn = "$tmpdir/index.json";
7001 my $cmd = "restore";
7003 my $param = [$pbs_backup_name, "index.json", $index_fn];
7004 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7005 my $index = PVE
::Tools
::file_get_contents
($index_fn);
7006 $index = decode_json
($index);
7008 foreach my $info (@{$index->{files
}}) {
7009 if ($info->{filename
} =~ m/^(drive-\S+).img.fidx$/) {
7011 if ($info->{size
} =~ m/^(\d+)$/) { # untaint size
7012 $devinfo->{$devname}->{size
} = $1;
7014 die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
7019 my $is_qemu_server_backup = scalar(
7020 grep { $_->{filename
} eq 'qemu-server.conf.blob' } @{$index->{files
}}
7022 if (!$is_qemu_server_backup) {
7023 die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
7025 my $has_firewall_config = scalar(grep { $_->{filename
} eq 'fw.conf.blob' } @{$index->{files
}});
7027 $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
7028 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7030 if ($has_firewall_config) {
7031 $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
7032 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7034 my $pve_firewall_dir = '/etc/pve/firewall';
7035 mkdir $pve_firewall_dir; # make sure the dir exists
7036 PVE
::Tools
::file_copy
($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
7039 my $fh = IO
::File-
>new($cfgfn, "r") ||
7040 die "unable to read qemu-server.conf - $!\n";
7042 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
7044 # fixme: rate limit?
7046 # create empty/temp config
7047 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\nlock: create");
7049 $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
7052 my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
7054 foreach my $virtdev (sort keys %$virtdev_hash) {
7055 my $d = $virtdev_hash->{$virtdev};
7056 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7058 # this fails if storage is unavailable
7059 my $volid = $d->{volid
};
7060 my $path = PVE
::Storage
::path
($storecfg, $volid);
7062 # for live-restore we only want to preload the efidisk and TPM state
7063 next if $options->{live
} && $virtdev ne 'efidisk0' && $virtdev ne 'tpmstate0';
7066 if (defined(my $ns = $scfg->{namespace
})) {
7067 @ns_arg = ('--ns', $ns);
7070 my $pbs_restore_cmd = [
7071 '/usr/bin/pbs-restore',
7072 '--repository', $repo,
7075 "$d->{devname}.img.fidx",
7080 push @$pbs_restore_cmd, '--format', $d->{format
} if $d->{format
};
7081 push @$pbs_restore_cmd, '--keyfile', $keyfile if -e
$keyfile;
7083 if (PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $volid)) {
7084 push @$pbs_restore_cmd, '--skip-zero';
7087 my $dbg_cmdstring = PVE
::Tools
::cmd2string
($pbs_restore_cmd);
7088 print "restore proxmox backup image: $dbg_cmdstring\n";
7089 run_command
($pbs_restore_cmd);
7092 $fh->seek(0, 0) || die "seek failed - $!\n";
7094 my $cookie = { netcount
=> 0 };
7095 while (defined(my $line = <$fh>)) {
7096 $new_conf_raw .= restore_update_config_line
(
7108 if ($err || !$options->{live
}) {
7109 $restore_deactivate_volumes->($storecfg, $virtdev_hash);
7115 $restore_destroy_volumes->($storecfg, $virtdev_hash);
7119 if ($options->{live
}) {
7120 # keep lock during live-restore
7121 $new_conf_raw .= "\nlock: create";
7124 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $options->{override_conf
});
7125 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7127 eval { rescan
($vmid, 1); };
7130 PVE
::AccessControl
::add_vm_to_pool
($vmid, $options->{pool
}) if $options->{pool
};
7132 if ($options->{live
}) {
7138 local $SIG{PIPE
} = sub { die "got signal ($!) - abort\n"; };
7140 my $conf = PVE
::QemuConfig-
>load_config($vmid);
7141 die "cannot do live-restore for template\n" if PVE
::QemuConfig-
>is_template($conf);
7143 # these special drives are already restored before start
7144 delete $devinfo->{'drive-efidisk0'};
7145 delete $devinfo->{'drive-tpmstate0-backup'};
7149 keyfile
=> $keyfile,
7150 snapshot
=> $pbs_backup_name,
7151 namespace
=> $namespace,
7153 pbs_live_restore
($vmid, $conf, $storecfg, $devinfo, $pbs_opts);
7155 PVE
::QemuConfig-
>remove_lock($vmid, "create");
7159 sub pbs_live_restore
{
7160 my ($vmid, $conf, $storecfg, $restored_disks, $opts) = @_;
7162 print "starting VM for live-restore\n";
7163 print "repository: '$opts->{repo}', snapshot: '$opts->{snapshot}'\n";
7165 my $pbs_backing = {};
7166 for my $ds (keys %$restored_disks) {
7167 $ds =~ m/^drive-(.*)$/;
7169 $pbs_backing->{$confname} = {
7170 repository
=> $opts->{repo
},
7171 snapshot
=> $opts->{snapshot
},
7172 archive
=> "$ds.img.fidx",
7174 $pbs_backing->{$confname}->{keyfile
} = $opts->{keyfile
} if -e
$opts->{keyfile
};
7175 $pbs_backing->{$confname}->{namespace
} = $opts->{namespace
} if defined($opts->{namespace
});
7177 my $drive = parse_drive
($confname, $conf->{$confname});
7178 print "restoring '$ds' to '$drive->{file}'\n";
7181 my $drives_streamed = 0;
7183 # make sure HA doesn't interrupt our restore by stopping the VM
7184 if (PVE
::HA
::Config
::vm_is_ha_managed
($vmid)) {
7185 run_command
(['ha-manager', 'set', "vm:$vmid", '--state', 'started']);
7188 # start VM with backing chain pointing to PBS backup, environment vars for PBS driver
7189 # in QEMU (PBS_PASSWORD and PBS_FINGERPRINT) are already set by our caller
7190 vm_start_nolock
($storecfg, $vmid, $conf, {paused
=> 1, 'pbs-backing' => $pbs_backing}, {});
7192 my $qmeventd_fd = register_qmeventd_handle
($vmid);
7194 # begin streaming, i.e. data copy from PBS to target disk for every vol,
7195 # this will effectively collapse the backing image chain consisting of
7196 # [target <- alloc-track -> PBS snapshot] to just [target] (alloc-track
7197 # removes itself once all backing images vanish with 'auto-remove=on')
7199 for my $ds (sort keys %$restored_disks) {
7200 my $job_id = "restore-$ds";
7201 mon_cmd
($vmid, 'block-stream',
7202 'job-id' => $job_id,
7205 $jobs->{$job_id} = {};
7208 mon_cmd
($vmid, 'cont');
7209 qemu_drive_mirror_monitor
($vmid, undef, $jobs, 'auto', 0, 'stream');
7211 print "restore-drive jobs finished successfully, removing all tracking block devices"
7212 ." to disconnect from Proxmox Backup Server\n";
7214 for my $ds (sort keys %$restored_disks) {
7215 mon_cmd
($vmid, 'blockdev-del', 'node-name' => "$ds-pbs");
7218 close($qmeventd_fd);
7224 warn "An error occurred during live-restore: $err\n";
7225 _do_vm_stop
($storecfg, $vmid, 1, 1, 10, 0, 1);
7226 die "live-restore failed\n";
7230 sub restore_vma_archive
{
7231 my ($archive, $vmid, $user, $opts, $comp) = @_;
7233 my $readfrom = $archive;
7235 my $cfg = PVE
::Storage
::config
();
7237 my $bwlimit = $opts->{bwlimit
};
7239 my $dbg_cmdstring = '';
7240 my $add_pipe = sub {
7242 push @$commands, $cmd;
7243 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
7244 $dbg_cmdstring .= PVE
::Tools
::cmd2string
($cmd);
7249 if ($archive eq '-') {
7252 # If we use a backup from a PVE defined storage we also consider that
7253 # storage's rate limit:
7254 my (undef, $volid) = PVE
::Storage
::path_to_volume_id
($cfg, $archive);
7255 if (defined($volid)) {
7256 my ($sid, undef) = PVE
::Storage
::parse_volume_id
($volid);
7257 my $readlimit = PVE
::Storage
::get_bandwidth_limit
('restore', [$sid], $bwlimit);
7259 print STDERR
"applying read rate limit: $readlimit\n";
7260 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
7261 $add_pipe->($cstream);
7267 my $info = PVE
::Storage
::decompressor_info
('vma', $comp);
7268 my $cmd = $info->{decompressor
};
7269 push @$cmd, $readfrom;
7273 my $tmpdir = "/var/tmp/vzdumptmp$$";
7276 # disable interrupts (always do cleanups)
7280 local $SIG{HUP
} = sub { warn "got interrupt - ignored\n"; };
7282 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
7283 POSIX
::mkfifo
($mapfifo, 0600);
7285 my $openfifo = sub { open($fifofh, '>', $mapfifo) or die $! };
7287 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
7292 my $devinfo = {}; # info about drives included in backup
7293 my $virtdev_hash = {}; # info about allocated drives
7295 my $rpcenv = PVE
::RPCEnvironment
::get
();
7297 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7299 # Note: $oldconf is undef if VM does not exist
7300 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7301 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7302 my $new_conf_raw = '';
7306 my $print_devmap = sub {
7307 my $cfgfn = "$tmpdir/qemu-server.conf";
7309 # we can read the config - that is already extracted
7310 my $fh = IO
::File-
>new($cfgfn, "r") ||
7311 die "unable to read qemu-server.conf - $!\n";
7313 my $fwcfgfn = "$tmpdir/qemu-server.fw";
7315 my $pve_firewall_dir = '/etc/pve/firewall';
7316 mkdir $pve_firewall_dir; # make sure the dir exists
7317 PVE
::Tools
::file_copy
($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
7320 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
7322 foreach my $info (values %{$virtdev_hash}) {
7323 my $storeid = $info->{storeid
};
7324 next if defined($storage_limits{$storeid});
7326 my $limit = PVE
::Storage
::get_bandwidth_limit
('restore', [$storeid], $bwlimit) // 0;
7327 print STDERR
"rate limit for storage $storeid: $limit KiB/s\n" if $limit;
7328 $storage_limits{$storeid} = $limit * 1024;
7331 foreach my $devname (keys %$devinfo) {
7332 die "found no device mapping information for device '$devname'\n"
7333 if !$devinfo->{$devname}->{virtdev
};
7336 # create empty/temp config
7338 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\n");
7339 $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
7343 my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
7345 # print restore information to $fifofh
7346 foreach my $virtdev (sort keys %$virtdev_hash) {
7347 my $d = $virtdev_hash->{$virtdev};
7348 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7350 my $storeid = $d->{storeid
};
7351 my $volid = $d->{volid
};
7354 if (my $limit = $storage_limits{$storeid}) {
7355 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
7358 my $write_zeros = 1;
7359 if (PVE
::Storage
::volume_has_feature
($cfg, 'sparseinit', $volid)) {
7363 my $path = PVE
::Storage
::path
($cfg, $volid);
7365 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
7367 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
7370 $fh->seek(0, 0) || die "seek failed - $!\n";
7372 my $cookie = { netcount
=> 0 };
7373 while (defined(my $line = <$fh>)) {
7374 $new_conf_raw .= restore_update_config_line
(
7391 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7392 local $SIG{ALRM
} = sub { die "got timeout\n"; };
7394 $oldtimeout = alarm($timeout);
7401 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
7402 my ($dev_id, $size, $devname) = ($1, $2, $3);
7403 $devinfo->{$devname} = { size
=> $size, dev_id
=> $dev_id };
7404 } elsif ($line =~ m/^CTIME: /) {
7405 # we correctly received the vma config, so we can disable
7406 # the timeout now for disk allocation (set to 10 minutes, so
7407 # that we always timeout if something goes wrong)
7410 print $fifofh "done\n";
7411 my $tmp = $oldtimeout || 0;
7412 $oldtimeout = undef;
7419 print "restore vma archive: $dbg_cmdstring\n";
7420 run_command
($commands, input
=> $input, outfunc
=> $parser, afterfork
=> $openfifo);
7424 alarm($oldtimeout) if $oldtimeout;
7426 $restore_deactivate_volumes->($cfg, $virtdev_hash);
7428 close($fifofh) if $fifofh;
7433 $restore_destroy_volumes->($cfg, $virtdev_hash);
7437 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $opts->{override_conf
});
7438 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7440 eval { rescan
($vmid, 1); };
7443 PVE
::AccessControl
::add_vm_to_pool
($vmid, $opts->{pool
}) if $opts->{pool
};
7446 sub restore_tar_archive
{
7447 my ($archive, $vmid, $user, $opts) = @_;
7449 if (scalar(keys $opts->{override_conf
}->%*) > 0) {
7450 my $keystring = join(' ', keys $opts->{override_conf
}->%*);
7451 die "cannot pass along options ($keystring) when restoring from tar archive\n";
7454 if ($archive ne '-') {
7455 my $firstfile = tar_archive_read_firstfile
($archive);
7456 die "ERROR: file '$archive' does not look like a QemuServer vzdump backup\n"
7457 if $firstfile ne 'qemu-server.conf';
7460 my $storecfg = PVE
::Storage
::config
();
7462 # avoid zombie disks when restoring over an existing VM -> cleanup first
7463 # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
7464 # skiplock=1 because qmrestore has set the 'create' lock itself already
7465 my $vmcfgfn = PVE
::QemuConfig-
>config_file($vmid);
7466 destroy_vm
($storecfg, $vmid, 1, { lock => 'restore' }) if -f
$vmcfgfn;
7468 my $tocmd = "/usr/lib/qemu-server/qmextract";
7470 $tocmd .= " --storage " . PVE
::Tools
::shellquote
($opts->{storage
}) if $opts->{storage
};
7471 $tocmd .= " --pool " . PVE
::Tools
::shellquote
($opts->{pool
}) if $opts->{pool
};
7472 $tocmd .= ' --prealloc' if $opts->{prealloc
};
7473 $tocmd .= ' --info' if $opts->{info
};
7475 # tar option "xf" does not autodetect compression when read from STDIN,
7476 # so we pipe to zcat
7477 my $cmd = "zcat -f|tar xf " . PVE
::Tools
::shellquote
($archive) . " " .
7478 PVE
::Tools
::shellquote
("--to-command=$tocmd");
7480 my $tmpdir = "/var/tmp/vzdumptmp$$";
7483 local $ENV{VZDUMP_TMPDIR
} = $tmpdir;
7484 local $ENV{VZDUMP_VMID
} = $vmid;
7485 local $ENV{VZDUMP_USER
} = $user;
7487 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7488 my $new_conf_raw = '';
7490 # disable interrupts (always do cleanups)
7494 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7502 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7504 if ($archive eq '-') {
7505 print "extracting archive from STDIN\n";
7506 run_command
($cmd, input
=> "<&STDIN");
7508 print "extracting archive '$archive'\n";
7512 return if $opts->{info
};
7516 my $statfile = "$tmpdir/qmrestore.stat";
7517 if (my $fd = IO
::File-
>new($statfile, "r")) {
7518 while (defined (my $line = <$fd>)) {
7519 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
7520 $map->{$1} = $2 if $1;
7522 print STDERR
"unable to parse line in statfile - $line\n";
7528 my $confsrc = "$tmpdir/qemu-server.conf";
7530 my $srcfd = IO
::File-
>new($confsrc, "r") || die "unable to open file '$confsrc'\n";
7532 my $cookie = { netcount
=> 0 };
7533 while (defined (my $line = <$srcfd>)) {
7534 $new_conf_raw .= restore_update_config_line
(
7545 tar_restore_cleanup
($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info
};
7551 PVE
::Tools
::file_set_contents
($conffile, $new_conf_raw);
7553 PVE
::Cluster
::cfs_update
(); # make sure we read new file
7555 eval { rescan
($vmid, 1); };
7559 sub foreach_storage_used_by_vm
{
7560 my ($conf, $func) = @_;
7564 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7565 my ($ds, $drive) = @_;
7566 return if drive_is_cdrom
($drive);
7568 my $volid = $drive->{file
};
7570 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
7571 $sidhash->{$sid} = $sid if $sid;
7574 foreach my $sid (sort keys %$sidhash) {
7579 my $qemu_snap_storage = {
7582 sub do_snapshots_with_qemu
{
7583 my ($storecfg, $volid, $deviceid) = @_;
7585 return if $deviceid =~ m/tpmstate0/;
7587 my $storage_name = PVE
::Storage
::parse_volume_id
($volid);
7588 my $scfg = $storecfg->{ids
}->{$storage_name};
7589 die "could not find storage '$storage_name'\n" if !defined($scfg);
7591 if ($qemu_snap_storage->{$scfg->{type
}} && !$scfg->{krbd
}){
7595 if ($volid =~ m/\.(qcow2|qed)$/){
7602 sub qga_check_running
{
7603 my ($vmid, $nowarn) = @_;
7605 eval { mon_cmd
($vmid, "guest-ping", timeout
=> 3); };
7607 warn "QEMU Guest Agent is not running - $@" if !$nowarn;
7613 sub template_create
{
7614 my ($vmid, $conf, $disk) = @_;
7616 my $storecfg = PVE
::Storage
::config
();
7618 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7619 my ($ds, $drive) = @_;
7621 return if drive_is_cdrom
($drive);
7622 return if $disk && $ds ne $disk;
7624 my $volid = $drive->{file
};
7625 return if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
7627 my $voliddst = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
7628 $drive->{file
} = $voliddst;
7629 $conf->{$ds} = print_drive
($drive);
7630 PVE
::QemuConfig-
>write_config($vmid, $conf);
7634 sub convert_iscsi_path
{
7637 if ($path =~ m
|^iscsi
://([^/]+)/([^/]+)/(.+)$|) {
7642 my $initiator_name = get_initiator_name
();
7644 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
7645 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
7648 die "cannot convert iscsi path '$path', unkown format\n";
7651 sub qemu_img_convert
{
7652 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
7654 my $storecfg = PVE
::Storage
::config
();
7655 my ($src_storeid, $src_volname) = PVE
::Storage
::parse_volume_id
($src_volid, 1);
7656 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid, 1);
7658 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
7662 my $src_is_iscsi = 0;
7666 PVE
::Storage
::activate_volumes
($storecfg, [$src_volid], $snapname);
7667 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7668 $src_format = qemu_img_format
($src_scfg, $src_volname);
7669 $src_path = PVE
::Storage
::path
($storecfg, $src_volid, $snapname);
7670 $src_is_iscsi = ($src_path =~ m
|^iscsi
://|);
7671 $cachemode = 'none' if $src_scfg->{type
} eq 'zfspool';
7672 } elsif (-f
$src_volid || -b
$src_volid) {
7673 $src_path = $src_volid;
7674 if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7679 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
7681 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7682 my $dst_format = qemu_img_format
($dst_scfg, $dst_volname);
7683 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7684 my $dst_is_iscsi = ($dst_path =~ m
|^iscsi
://|);
7687 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
7688 push @$cmd, '-l', "snapshot.name=$snapname"
7689 if $snapname && $src_format && $src_format eq "qcow2";
7690 push @$cmd, '-t', 'none' if $dst_scfg->{type
} eq 'zfspool';
7691 push @$cmd, '-T', $cachemode if defined($cachemode);
7693 if ($src_is_iscsi) {
7694 push @$cmd, '--image-opts';
7695 $src_path = convert_iscsi_path
($src_path);
7696 } elsif ($src_format) {
7697 push @$cmd, '-f', $src_format;
7700 if ($dst_is_iscsi) {
7701 push @$cmd, '--target-image-opts';
7702 $dst_path = convert_iscsi_path
($dst_path);
7704 push @$cmd, '-O', $dst_format;
7707 push @$cmd, $src_path;
7709 if (!$dst_is_iscsi && $is_zero_initialized) {
7710 push @$cmd, "zeroinit:$dst_path";
7712 push @$cmd, $dst_path;
7717 if($line =~ m/\((\S+)\/100\
%\)/){
7719 my $transferred = int($size * $percent / 100);
7720 my $total_h = render_bytes
($size, 1);
7721 my $transferred_h = render_bytes
($transferred, 1);
7723 print "transferred $transferred_h of $total_h ($percent%)\n";
7728 eval { run_command
($cmd, timeout
=> undef, outfunc
=> $parser); };
7730 die "copy failed: $err" if $err;
7733 sub qemu_img_format
{
7734 my ($scfg, $volname) = @_;
7736 if ($scfg->{path
} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7743 sub qemu_drive_mirror
{
7744 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
7746 $jobs = {} if !$jobs;
7750 $jobs->{"drive-$drive"} = {};
7752 if ($dst_volid =~ /^nbd:/) {
7753 $qemu_target = $dst_volid;
7756 my $storecfg = PVE
::Storage
::config
();
7757 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid);
7759 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7761 $format = qemu_img_format
($dst_scfg, $dst_volname);
7763 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7765 $qemu_target = $is_zero_initialized ?
"zeroinit:$dst_path" : $dst_path;
7768 my $opts = { timeout
=> 10, device
=> "drive-$drive", mode
=> "existing", sync
=> "full", target
=> $qemu_target };
7769 $opts->{format
} = $format if $format;
7771 if (defined($src_bitmap)) {
7772 $opts->{sync
} = 'incremental';
7773 $opts->{bitmap
} = $src_bitmap;
7774 print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
7777 if (defined($bwlimit)) {
7778 $opts->{speed
} = $bwlimit * 1024;
7779 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7781 print "drive mirror is starting for drive-$drive\n";
7784 # if a job already runs for this device we get an error, catch it for cleanup
7785 eval { mon_cmd
($vmid, "drive-mirror", %$opts); };
7787 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7789 die "mirroring error: $err\n";
7792 qemu_drive_mirror_monitor
($vmid, $vmiddst, $jobs, $completion, $qga);
7795 # $completion can be either
7796 # 'complete': wait until all jobs are ready, block-job-complete them (default)
7797 # 'cancel': wait until all jobs are ready, block-job-cancel them
7798 # 'skip': wait until all jobs are ready, return with block jobs in ready state
7799 # 'auto': wait until all jobs disappear, only use for jobs which complete automatically
7800 sub qemu_drive_mirror_monitor
{
7801 my ($vmid, $vmiddst, $jobs, $completion, $qga, $op) = @_;
7803 $completion //= 'complete';
7807 my $err_complete = 0;
7809 my $starttime = time ();
7811 die "block job ('$op') timed out\n" if $err_complete > 300;
7813 my $stats = mon_cmd
($vmid, "query-block-jobs");
7816 my $running_jobs = {};
7817 for my $stat (@$stats) {
7818 next if $stat->{type
} ne $op;
7819 $running_jobs->{$stat->{device
}} = $stat;
7822 my $readycounter = 0;
7824 for my $job_id (sort keys %$jobs) {
7825 my $job = $running_jobs->{$job_id};
7827 my $vanished = !defined($job);
7828 my $complete = defined($jobs->{$job_id}->{complete
}) && $vanished;
7829 if($complete || ($vanished && $completion eq 'auto')) {
7830 print "$job_id: $op-job finished\n";
7831 delete $jobs->{$job_id};
7835 die "$job_id: '$op' has been cancelled\n" if !defined($job);
7837 my $busy = $job->{busy
};
7838 my $ready = $job->{ready
};
7839 if (my $total = $job->{len
}) {
7840 my $transferred = $job->{offset
} || 0;
7841 my $remaining = $total - $transferred;
7842 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7844 my $duration = $ctime - $starttime;
7845 my $total_h = render_bytes
($total, 1);
7846 my $transferred_h = render_bytes
($transferred, 1);
7848 my $status = sprintf(
7849 "transferred $transferred_h of $total_h ($percent%%) in %s",
7850 render_duration
($duration),
7855 $status .= ", still busy"; # shouldn't even happen? but mirror is weird
7857 $status .= ", ready";
7860 print "$job_id: $status\n" if !$jobs->{$job_id}->{ready
};
7861 $jobs->{$job_id}->{ready
} = $ready;
7864 $readycounter++ if $job->{ready
};
7867 last if scalar(keys %$jobs) == 0;
7869 if ($readycounter == scalar(keys %$jobs)) {
7870 print "all '$op' jobs are ready\n";
7872 # do the complete later (or has already been done)
7873 last if $completion eq 'skip' || $completion eq 'auto';
7875 if ($vmiddst && $vmiddst != $vmid) {
7876 my $agent_running = $qga && qga_check_running
($vmid);
7877 if ($agent_running) {
7878 print "freeze filesystem\n";
7879 eval { mon_cmd
($vmid, "guest-fsfreeze-freeze"); };
7882 print "suspend vm\n";
7883 eval { PVE
::QemuServer
::vm_suspend
($vmid, 1); };
7887 # if we clone a disk for a new target vm, we don't switch the disk
7888 PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs);
7890 if ($agent_running) {
7891 print "unfreeze filesystem\n";
7892 eval { mon_cmd
($vmid, "guest-fsfreeze-thaw"); };
7895 print "resume vm\n";
7896 eval { PVE
::QemuServer
::vm_resume
($vmid, 1, 1); };
7903 for my $job_id (sort keys %$jobs) {
7904 # try to switch the disk if source and destination are on the same guest
7905 print "$job_id: Completing block job_id...\n";
7908 if ($completion eq 'complete') {
7909 $op = 'block-job-complete';
7910 } elsif ($completion eq 'cancel') {
7911 $op = 'block-job-cancel';
7913 die "invalid completion value: $completion\n";
7915 eval { mon_cmd
($vmid, $op, device
=> $job_id) };
7916 if ($@ =~ m/cannot be completed/) {
7917 print "$job_id: block job cannot be completed, trying again.\n";
7920 print "$job_id: Completed successfully.\n";
7921 $jobs->{$job_id}->{complete
} = 1;
7932 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7933 die "block job ($op) error: $err";
7937 sub qemu_blockjobs_cancel
{
7938 my ($vmid, $jobs) = @_;
7940 foreach my $job (keys %$jobs) {
7941 print "$job: Cancelling block job\n";
7942 eval { mon_cmd
($vmid, "block-job-cancel", device
=> $job); };
7943 $jobs->{$job}->{cancel
} = 1;
7947 my $stats = mon_cmd
($vmid, "query-block-jobs");
7949 my $running_jobs = {};
7950 foreach my $stat (@$stats) {
7951 $running_jobs->{$stat->{device
}} = $stat;
7954 foreach my $job (keys %$jobs) {
7956 if (defined($jobs->{$job}->{cancel
}) && !defined($running_jobs->{$job})) {
7957 print "$job: Done.\n";
7958 delete $jobs->{$job};
7962 last if scalar(keys %$jobs) == 0;
7968 # Check for bug #4525: drive-mirror will open the target drive with the same aio setting as the
7969 # source, but some storages have problems with io_uring, sometimes even leading to crashes.
7970 my sub clone_disk_check_io_uring
{
7971 my ($src_drive, $storecfg, $src_storeid, $dst_storeid, $use_drive_mirror) = @_;
7973 return if !$use_drive_mirror;
7975 # Don't complain when not changing storage.
7976 # Assume if it works for the source, it'll work for the target too.
7977 return if $src_storeid eq $dst_storeid;
7979 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7980 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7982 my $cache_direct = drive_uses_cache_direct
($src_drive);
7984 my $src_uses_io_uring;
7985 if ($src_drive->{aio
}) {
7986 $src_uses_io_uring = $src_drive->{aio
} eq 'io_uring';
7988 $src_uses_io_uring = storage_allows_io_uring_default
($src_scfg, $cache_direct);
7991 die "target storage is known to cause issues with aio=io_uring (used by current drive)\n"
7992 if $src_uses_io_uring && !storage_allows_io_uring_default
($dst_scfg, $cache_direct);
7996 my ($storecfg, $source, $dest, $full, $newvollist, $jobs, $completion, $qga, $bwlimit) = @_;
7998 my ($vmid, $running) = $source->@{qw(vmid running)};
7999 my ($src_drivename, $drive, $snapname) = $source->@{qw(drivename drive snapname)};
8001 my ($newvmid, $dst_drivename, $efisize) = $dest->@{qw(vmid drivename efisize)};
8002 my ($storage, $format) = $dest->@{qw(storage format)};
8004 my $use_drive_mirror = $full && $running && $src_drivename && !$snapname;
8006 if ($src_drivename && $dst_drivename && $src_drivename ne $dst_drivename) {
8007 die "cloning from/to EFI disk requires EFI disk\n"
8008 if $src_drivename eq 'efidisk0' || $dst_drivename eq 'efidisk0';
8009 die "cloning from/to TPM state requires TPM state\n"
8010 if $src_drivename eq 'tpmstate0' || $dst_drivename eq 'tpmstate0';
8012 # This would lead to two device nodes in QEMU pointing to the same backing image!
8013 die "cannot change drive name when cloning disk from/to the same VM\n"
8014 if $use_drive_mirror && $vmid == $newvmid;
8017 die "cannot move TPM state while VM is running\n"
8018 if $use_drive_mirror && $src_drivename eq 'tpmstate0';
8022 print "create " . ($full ?
'full' : 'linked') . " clone of drive ";
8023 print "$src_drivename " if $src_drivename;
8024 print "($drive->{file})\n";
8027 $newvolid = PVE
::Storage
::vdisk_clone
($storecfg, $drive->{file
}, $newvmid, $snapname);
8028 push @$newvollist, $newvolid;
8030 my ($src_storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
8031 my $storeid = $storage || $src_storeid;
8033 my $dst_format = resolve_dst_disk_format
($storecfg, $storeid, $volname, $format);
8037 if (drive_is_cloudinit
($drive)) {
8038 $name = "vm-$newvmid-cloudinit";
8039 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8040 if ($scfg->{path
}) {
8041 $name .= ".$dst_format";
8044 $size = PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
;
8045 } elsif ($dst_drivename eq 'efidisk0') {
8046 $size = $efisize or die "internal error - need to specify EFI disk size\n";
8047 } elsif ($dst_drivename eq 'tpmstate0') {
8048 $dst_format = 'raw';
8049 $size = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8051 clone_disk_check_io_uring
($drive, $storecfg, $src_storeid, $storeid, $use_drive_mirror);
8053 ($size) = PVE
::Storage
::volume_size_info
($storecfg, $drive->{file
}, 10);
8055 $newvolid = PVE
::Storage
::vdisk_alloc
(
8056 $storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024)
8058 push @$newvollist, $newvolid;
8060 PVE
::Storage
::activate_volumes
($storecfg, [$newvolid]);
8062 if (drive_is_cloudinit
($drive)) {
8063 # when cloning multiple disks (e.g. during clone_vm) it might be the last disk
8064 # if this is the case, we have to complete any block-jobs still there from
8065 # previous drive-mirrors
8066 if (($completion eq 'complete') && (scalar(keys %$jobs) > 0)) {
8067 qemu_drive_mirror_monitor
($vmid, $newvmid, $jobs, $completion, $qga);
8072 my $sparseinit = PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $newvolid);
8073 if ($use_drive_mirror) {
8074 qemu_drive_mirror
($vmid, $src_drivename, $newvolid, $newvmid, $sparseinit, $jobs,
8075 $completion, $qga, $bwlimit);
8077 # TODO: handle bwlimits
8078 if ($dst_drivename eq 'efidisk0') {
8079 # the relevant data on the efidisk may be smaller than the source
8080 # e.g. on RBD/ZFS, so we use dd to copy only the amount
8081 # that is given by the OVMF_VARS.fd
8082 my $src_path = PVE
::Storage
::path
($storecfg, $drive->{file
}, $snapname);
8083 my $dst_path = PVE
::Storage
::path
($storecfg, $newvolid);
8085 my $src_format = (PVE
::Storage
::parse_volname
($storecfg, $drive->{file
}))[6];
8087 # better for Ceph if block size is not too small, see bug #3324
8090 my $cmd = ['qemu-img', 'dd', '-n', '-O', $dst_format];
8092 if ($src_format eq 'qcow2' && $snapname) {
8093 die "cannot clone qcow2 EFI disk snapshot - requires QEMU >= 6.2\n"
8094 if !min_version
(kvm_user_version
(), 6, 2);
8095 push $cmd->@*, '-l', $snapname;
8097 push $cmd->@*, "bs=$bs", "osize=$size", "if=$src_path", "of=$dst_path";
8100 qemu_img_convert
($drive->{file
}, $newvolid, $size, $snapname, $sparseinit);
8106 my ($size) = eval { PVE
::Storage
::volume_size_info
($storecfg, $newvolid, 10) };
8108 my $disk = dclone
($drive);
8109 delete $disk->{format
};
8110 $disk->{file
} = $newvolid;
8111 $disk->{size
} = $size if defined($size);
8116 sub get_running_qemu_version
{
8118 my $res = mon_cmd
($vmid, "query-version");
8119 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
8122 sub qemu_use_old_bios_files
{
8123 my ($machine_type) = @_;
8125 return if !$machine_type;
8127 my $use_old_bios_files = undef;
8129 if ($machine_type =~ m/^(\S+)\.pxe$/) {
8131 $use_old_bios_files = 1;
8133 my $version = extract_version
($machine_type, kvm_user_version
());
8134 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
8135 # load new efi bios files on migration. So this hack is required to allow
8136 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
8137 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
8138 $use_old_bios_files = !min_version
($version, 2, 4);
8141 return ($use_old_bios_files, $machine_type);
8144 sub get_efivars_size
{
8145 my ($conf, $efidisk) = @_;
8147 my $arch = get_vm_arch
($conf);
8148 $efidisk //= $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
8149 my $smm = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
8150 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8151 return -s
$ovmf_vars;
8154 sub update_efidisk_size
{
8157 return if !defined($conf->{efidisk0
});
8159 my $disk = PVE
::QemuServer
::parse_drive
('efidisk0', $conf->{efidisk0
});
8160 $disk->{size
} = get_efivars_size
($conf);
8161 $conf->{efidisk0
} = print_drive
($disk);
8166 sub update_tpmstate_size
{
8169 my $disk = PVE
::QemuServer
::parse_drive
('tpmstate0', $conf->{tpmstate0
});
8170 $disk->{size
} = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8171 $conf->{tpmstate0
} = print_drive
($disk);
8174 sub create_efidisk
($$$$$$$) {
8175 my ($storecfg, $storeid, $vmid, $fmt, $arch, $efidisk, $smm) = @_;
8177 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8179 my $vars_size_b = -s
$ovmf_vars;
8180 my $vars_size = PVE
::Tools
::convert_size
($vars_size_b, 'b' => 'kb');
8181 my $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
8182 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
8184 qemu_img_convert
($ovmf_vars, $volid, $vars_size_b, undef, 0);
8185 my ($size) = PVE
::Storage
::volume_size_info
($storecfg, $volid, 3);
8187 return ($volid, $size/1024);
8190 sub vm_iothreads_list
{
8193 my $res = mon_cmd
($vmid, 'query-iothreads');
8196 foreach my $iothread (@$res) {
8197 $iothreads->{ $iothread->{id
} } = $iothread->{"thread-id"};
8204 my ($conf, $drive) = @_;
8208 if (!$conf->{scsihw
} || ($conf->{scsihw
} =~ m/^lsi/)) {
8210 } elsif ($conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single')) {
8216 my $controller = int($drive->{index} / $maxdev);
8217 my $controller_prefix = ($conf->{scsihw
} && $conf->{scsihw
} eq 'virtio-scsi-single')
8221 return ($maxdev, $controller, $controller_prefix);
8224 sub resolve_dst_disk_format
{
8225 my ($storecfg, $storeid, $src_volname, $format) = @_;
8226 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
8229 # if no target format is specified, use the source disk format as hint
8231 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8232 $format = qemu_img_format
($scfg, $src_volname);
8238 # test if requested format is supported - else use default
8239 my $supported = grep { $_ eq $format } @$validFormats;
8240 $format = $defFormat if !$supported;
8244 # NOTE: if this logic changes, please update docs & possibly gui logic
8245 sub find_vmstate_storage
{
8246 my ($conf, $storecfg) = @_;
8248 # first, return storage from conf if set
8249 return $conf->{vmstatestorage
} if $conf->{vmstatestorage
};
8251 my ($target, $shared, $local);
8253 foreach_storage_used_by_vm
($conf, sub {
8255 my $scfg = PVE
::Storage
::storage_config
($storecfg, $sid);
8256 my $dst = $scfg->{shared
} ? \
$shared : \
$local;
8257 $$dst = $sid if !$$dst || $scfg->{path
}; # prefer file based storage
8260 # second, use shared storage where VM has at least one disk
8261 # third, use local storage where VM has at least one disk
8262 # fall back to local storage
8263 $target = $shared // $local // 'local';
8269 my ($uuid, $uuid_str);
8270 UUID
::generate
($uuid);
8271 UUID
::unparse
($uuid, $uuid_str);
8275 sub generate_smbios1_uuid
{
8276 return "uuid=".generate_uuid
();
8282 mon_cmd
($vmid, 'nbd-server-stop');
8285 sub create_reboot_request
{
8287 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
8288 or die "failed to create reboot trigger file: $!\n";
8292 sub clear_reboot_request
{
8294 my $path = "/run/qemu-server/$vmid.reboot";
8297 $res = unlink($path);
8298 die "could not remove reboot request for $vmid: $!"
8299 if !$res && $! != POSIX
::ENOENT
;
8304 sub bootorder_from_legacy
{
8305 my ($conf, $bootcfg) = @_;
8307 my $boot = $bootcfg->{legacy
} || $boot_fmt->{legacy
}->{default};
8308 my $bootindex_hash = {};
8310 foreach my $o (split(//, $boot)) {
8311 $bootindex_hash->{$o} = $i*100;
8317 PVE
::QemuConfig-
>foreach_volume($conf, sub {
8318 my ($ds, $drive) = @_;
8320 if (drive_is_cdrom
($drive, 1)) {
8321 if ($bootindex_hash->{d
}) {
8322 $bootorder->{$ds} = $bootindex_hash->{d
};
8323 $bootindex_hash->{d
} += 1;
8325 } elsif ($bootindex_hash->{c
}) {
8326 $bootorder->{$ds} = $bootindex_hash->{c
}
8327 if $conf->{bootdisk
} && $conf->{bootdisk
} eq $ds;
8328 $bootindex_hash->{c
} += 1;
8332 if ($bootindex_hash->{n
}) {
8333 for (my $i = 0; $i < $MAX_NETS; $i++) {
8334 my $netname = "net$i";
8335 next if !$conf->{$netname};
8336 $bootorder->{$netname} = $bootindex_hash->{n
};
8337 $bootindex_hash->{n
} += 1;
8344 # Generate default device list for 'boot: order=' property. Matches legacy
8345 # default boot order, but with explicit device names. This is important, since
8346 # the fallback for when neither 'order' nor the old format is specified relies
8347 # on 'bootorder_from_legacy' above, and it would be confusing if this diverges.
8348 sub get_default_bootdevices
{
8354 my $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 0);
8355 push @ret, $first if $first;
8358 $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 1);
8359 push @ret, $first if $first;
8362 for (my $i = 0; $i < $MAX_NETS; $i++) {
8363 my $netname = "net$i";
8364 next if !$conf->{$netname};
8365 push @ret, $netname;
8372 sub device_bootorder
{
8375 return bootorder_from_legacy
($conf) if !defined($conf->{boot
});
8377 my $boot = parse_property_string
($boot_fmt, $conf->{boot
});
8380 if (!defined($boot) || $boot->{legacy
}) {
8381 $bootorder = bootorder_from_legacy
($conf, $boot);
8382 } elsif ($boot->{order
}) {
8383 my $i = 100; # start at 100 to allow user to insert devices before us with -args
8384 for my $dev (PVE
::Tools
::split_list
($boot->{order
})) {
8385 $bootorder->{$dev} = $i++;
8392 sub register_qmeventd_handle
{
8396 my $peer = "/var/run/qmeventd.sock";
8401 $fh = IO
::Socket
::UNIX-
>new(Peer
=> $peer, Blocking
=> 0, Timeout
=> 1);
8403 if ($! != EINTR
&& $! != EAGAIN
) {
8404 die "unable to connect to qmeventd socket (vmid: $vmid) - $!\n";
8407 die "unable to connect to qmeventd socket (vmid: $vmid) - timeout "
8408 . "after $count retries\n";
8413 # send handshake to mark VM as backing up
8414 print $fh to_json
({vzdump
=> {vmid
=> "$vmid"}});
8416 # return handle to be closed later when inhibit is no longer required
8420 # bash completion helper
8422 sub complete_backup_archives
{
8423 my ($cmdname, $pname, $cvalue) = @_;
8425 my $cfg = PVE
::Storage
::config
();
8429 if ($cvalue =~ m/^([^:]+):/) {
8433 my $data = PVE
::Storage
::template_list
($cfg, $storeid, 'backup');
8436 foreach my $id (keys %$data) {
8437 foreach my $item (@{$data->{$id}}) {
8438 next if $item->{format
} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
8439 push @$res, $item->{volid
} if defined($item->{volid
});
8446 my $complete_vmid_full = sub {
8449 my $idlist = vmstatus
();
8453 foreach my $id (keys %$idlist) {
8454 my $d = $idlist->{$id};
8455 if (defined($running)) {
8456 next if $d->{template
};
8457 next if $running && $d->{status
} ne 'running';
8458 next if !$running && $d->{status
} eq 'running';
8467 return &$complete_vmid_full();
8470 sub complete_vmid_stopped
{
8471 return &$complete_vmid_full(0);
8474 sub complete_vmid_running
{
8475 return &$complete_vmid_full(1);
8478 sub complete_storage
{
8480 my $cfg = PVE
::Storage
::config
();
8481 my $ids = $cfg->{ids
};
8484 foreach my $sid (keys %$ids) {
8485 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, undef, 1);
8486 next if !$ids->{$sid}->{content
}->{images
};
8493 sub complete_migration_storage
{
8494 my ($cmd, $param, $current_value, $all_args) = @_;
8496 my $targetnode = @$all_args[1];
8498 my $cfg = PVE
::Storage
::config
();
8499 my $ids = $cfg->{ids
};
8502 foreach my $sid (keys %$ids) {
8503 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, $targetnode, 1);
8504 next if !$ids->{$sid}->{content
}->{images
};
8513 my $qmpstatus = eval {
8514 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid);
8515 mon_cmd
($vmid, "query-status");
8518 return $qmpstatus && $qmpstatus->{status
} eq "paused";
8521 sub check_volume_storage_type
{
8522 my ($storecfg, $vol) = @_;
8524 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($vol);
8525 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8526 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $vol);
8528 die "storage '$storeid' does not support content-type '$vtype'\n"
8529 if !$scfg->{content
}->{$vtype};
8534 sub add_nets_bridge_fdb
{
8535 my ($conf, $vmid) = @_;
8537 for my $opt (keys %$conf) {
8538 next if $opt !~ m/^net(\d+)$/;
8539 my $iface = "tap${vmid}i$1";
8540 # NOTE: expect setups with learning off to *not* use auto-random-generation of MAC on start
8541 my $net = parse_net
($conf->{$opt}, 1) or next;
8543 my $mac = $net->{macaddr
};
8545 log_warn
("MAC learning disabled, but vNIC '$iface' has no static MAC to add to forwarding DB!")
8546 if !file_read_firstline
("/sys/class/net/$iface/brport/learning");
8550 my $bridge = $net->{bridge
};
8552 PVE
::Network
::SDN
::Zones
::add_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8553 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8554 PVE
::Network
::add_bridge_fdb
($iface, $mac, $net->{firewall
});
8559 sub del_nets_bridge_fdb
{
8560 my ($conf, $vmid) = @_;
8562 for my $opt (keys %$conf) {
8563 next if $opt !~ m/^net(\d+)$/;
8564 my $iface = "tap${vmid}i$1";
8566 my $net = parse_net
($conf->{$opt}) or next;
8567 my $mac = $net->{macaddr
} or next;
8569 my $bridge = $net->{bridge
};
8571 PVE
::Network
::SDN
::Zones
::del_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8572 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8573 PVE
::Network
::del_bridge_fdb
($iface, $mac, $net->{firewall
});