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 - 5.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 $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1601 sub print_drive_commandline_full
{
1602 my ($storecfg, $vmid, $drive, $pbs_name, $io_uring) = @_;
1605 my $volid = $drive->{file
};
1606 my $format = $drive->{format
};
1607 my $drive_id = get_drive_id
($drive);
1609 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1610 my $scfg = $storeid ? PVE
::Storage
::storage_config
($storecfg, $storeid) : undef;
1612 if (drive_is_cdrom
($drive)) {
1613 $path = get_iso_path
($storecfg, $vmid, $volid);
1614 die "$drive_id: cannot back cdrom drive with PBS snapshot\n" if $pbs_name;
1617 $path = PVE
::Storage
::path
($storecfg, $volid);
1618 $format //= qemu_img_format
($scfg, $volname);
1625 my $is_rbd = $path =~ m/^rbd:/;
1628 my @qemu_drive_options = qw(heads secs cyls trans media cache rerror werror aio discard);
1629 foreach my $o (@qemu_drive_options) {
1630 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1633 # snapshot only accepts on|off
1634 if (defined($drive->{snapshot
})) {
1635 my $v = $drive->{snapshot
} ?
'on' : 'off';
1636 $opts .= ",snapshot=$v";
1639 if (defined($drive->{ro
})) { # ro maps to QEMUs `readonly`, which accepts `on` or `off` only
1640 $opts .= ",readonly=" . ($drive->{ro
} ?
'on' : 'off');
1643 foreach my $type (['', '-total'], [_rd
=> '-read'], [_wr
=> '-write']) {
1644 my ($dir, $qmpname) = @$type;
1645 if (my $v = $drive->{"mbps$dir"}) {
1646 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1648 if (my $v = $drive->{"mbps${dir}_max"}) {
1649 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1651 if (my $v = $drive->{"bps${dir}_max_length"}) {
1652 $opts .= ",throttling.bps$qmpname-max-length=$v";
1654 if (my $v = $drive->{"iops${dir}"}) {
1655 $opts .= ",throttling.iops$qmpname=$v";
1657 if (my $v = $drive->{"iops${dir}_max"}) {
1658 $opts .= ",throttling.iops$qmpname-max=$v";
1660 if (my $v = $drive->{"iops${dir}_max_length"}) {
1661 $opts .= ",throttling.iops$qmpname-max-length=$v";
1666 $format = "rbd" if $is_rbd;
1667 die "$drive_id: Proxmox Backup Server backed drive cannot auto-detect the format\n"
1669 $opts .= ",format=alloc-track,file.driver=$format";
1671 $opts .= ",format=$format";
1674 my $cache_direct = 0;
1676 if (my $cache = $drive->{cache
}) {
1677 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1678 } elsif (!drive_is_cdrom
($drive) && !($scfg && $scfg->{type
} eq 'btrfs' && !$scfg->{nocow
})) {
1679 $opts .= ",cache=none";
1683 # io_uring with cache mode writeback or writethrough on krbd will hang...
1684 my $rbd_no_io_uring = $scfg && $scfg->{type
} eq 'rbd' && $scfg->{krbd
} && !$cache_direct;
1686 # io_uring with cache mode writeback or writethrough on LVM will hang, without cache only
1687 # sometimes, just plain disable...
1688 my $lvm_no_io_uring = $scfg && $scfg->{type
} eq 'lvm';
1690 # io_uring causes problems when used with CIFS since kernel 5.15
1691 # Some discussion: https://www.spinics.net/lists/linux-cifs/msg26734.html
1692 my $cifs_no_io_uring = $scfg && $scfg->{type
} eq 'cifs';
1694 if (!$drive->{aio
}) {
1695 if ($io_uring && !$rbd_no_io_uring && !$lvm_no_io_uring && !$cifs_no_io_uring) {
1696 # io_uring supports all cache modes
1697 $opts .= ",aio=io_uring";
1699 # aio native works only with O_DIRECT
1701 $opts .= ",aio=native";
1703 $opts .= ",aio=threads";
1708 if (!drive_is_cdrom
($drive)) {
1710 if (defined($drive->{detect_zeroes
}) && !$drive->{detect_zeroes
}) {
1711 $detectzeroes = 'off';
1712 } elsif ($drive->{discard
}) {
1713 $detectzeroes = $drive->{discard
} eq 'on' ?
'unmap' : 'on';
1715 # This used to be our default with discard not being specified:
1716 $detectzeroes = 'on';
1719 # note: 'detect-zeroes' works per blockdev and we want it to persist
1720 # after the alloc-track is removed, so put it on 'file' directly
1721 my $dz_param = $pbs_name ?
"file.detect-zeroes" : "detect-zeroes";
1722 $opts .= ",$dz_param=$detectzeroes" if $detectzeroes;
1726 $opts .= ",backing=$pbs_name";
1727 $opts .= ",auto-remove=on";
1730 # my $file_param = $pbs_name ? "file.file.filename" : "file";
1731 my $file_param = "file";
1733 # non-rbd drivers require the underlying file to be a seperate block
1734 # node, so add a second .file indirection
1735 $file_param .= ".file" if !$is_rbd;
1736 $file_param .= ".filename";
1738 my $pathinfo = $path ?
"$file_param=$path," : '';
1740 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1743 sub print_pbs_blockdev
{
1744 my ($pbs_conf, $pbs_name) = @_;
1745 my $blockdev = "driver=pbs,node-name=$pbs_name,read-only=on";
1746 $blockdev .= ",repository=$pbs_conf->{repository}";
1747 $blockdev .= ",namespace=$pbs_conf->{namespace}" if $pbs_conf->{namespace
};
1748 $blockdev .= ",snapshot=$pbs_conf->{snapshot}";
1749 $blockdev .= ",archive=$pbs_conf->{archive}";
1750 $blockdev .= ",keyfile=$pbs_conf->{keyfile}" if $pbs_conf->{keyfile
};
1754 sub print_netdevice_full
{
1755 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version) = @_;
1757 my $device = $net->{model
};
1758 if ($net->{model
} eq 'virtio') {
1759 $device = 'virtio-net-pci';
1762 my $pciaddr = print_pci_addr
("$netid", $bridges, $arch, $machine_type);
1763 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1764 if ($net->{queues
} && $net->{queues
} > 1 && $net->{model
} eq 'virtio'){
1765 # Consider we have N queues, the number of vectors needed is 2 * N + 2, i.e., one per in
1766 # and out of each queue plus one config interrupt and control vector queue
1767 my $vectors = $net->{queues
} * 2 + 2;
1768 $tmpstr .= ",vectors=$vectors,mq=on";
1769 if (min_version
($machine_version, 7, 1)) {
1770 $tmpstr .= ",packed=on";
1774 if (min_version
($machine_version, 7, 1) && $net->{model
} eq 'virtio'){
1775 $tmpstr .= ",rx_queue_size=1024,tx_queue_size=1024";
1778 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex
} ;
1780 if (my $mtu = $net->{mtu
}) {
1781 if ($net->{model
} eq 'virtio' && $net->{bridge
}) {
1782 my $bridge_mtu = PVE
::Network
::read_bridge_mtu
($net->{bridge
});
1785 } elsif ($mtu < 576) {
1786 die "netdev $netid: MTU '$mtu' is smaller than the IP minimum MTU '576'\n";
1787 } elsif ($mtu > $bridge_mtu) {
1788 die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
1790 $tmpstr .= ",host_mtu=$mtu";
1792 warn "WARN: netdev $netid: ignoring MTU '$mtu', not using VirtIO or no bridge configured.\n";
1796 if ($use_old_bios_files) {
1798 if ($device eq 'virtio-net-pci') {
1799 $romfile = 'pxe-virtio.rom';
1800 } elsif ($device eq 'e1000') {
1801 $romfile = 'pxe-e1000.rom';
1802 } elsif ($device eq 'e1000e') {
1803 $romfile = 'pxe-e1000e.rom';
1804 } elsif ($device eq 'ne2k') {
1805 $romfile = 'pxe-ne2k_pci.rom';
1806 } elsif ($device eq 'pcnet') {
1807 $romfile = 'pxe-pcnet.rom';
1808 } elsif ($device eq 'rtl8139') {
1809 $romfile = 'pxe-rtl8139.rom';
1811 $tmpstr .= ",romfile=$romfile" if $romfile;
1817 sub print_netdev_full
{
1818 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
1821 if ($netid =~ m/^net(\d+)$/) {
1825 die "got strange net id '$i'\n" if $i >= ${MAX_NETS
};
1827 my $ifname = "tap${vmid}i$i";
1829 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1830 die "interface name '$ifname' is too long (max 15 character)\n"
1831 if length($ifname) >= 16;
1833 my $vhostparam = '';
1834 if (is_native
($arch)) {
1835 $vhostparam = ',vhost=on' if kernel_has_vhost_net
() && $net->{model
} eq 'virtio';
1838 my $vmname = $conf->{name
} || "vm$vmid";
1841 my $script = $hotplug ?
"pve-bridge-hotplug" : "pve-bridge";
1843 if ($net->{bridge
}) {
1844 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script"
1845 .",downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1847 $netdev = "type=user,id=$netid,hostname=$vmname";
1850 $netdev .= ",queues=$net->{queues}" if ($net->{queues
} && $net->{model
} eq 'virtio');
1856 'cirrus' => 'cirrus-vga',
1858 'vmware' => 'vmware-svga',
1859 'virtio' => 'virtio-vga',
1860 'virtio-gl' => 'virtio-vga-gl',
1863 sub print_vga_device
{
1864 my ($conf, $vga, $arch, $machine_version, $machine, $id, $qxlnum, $bridges) = @_;
1866 my $type = $vga_map->{$vga->{type
}};
1867 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
1868 $type = 'virtio-gpu';
1870 my $vgamem_mb = $vga->{memory
};
1872 my $max_outputs = '';
1874 $type = $id ?
'qxl' : 'qxl-vga';
1876 if (!$conf->{ostype
} || $conf->{ostype
} =~ m/^(?:l\d\d)|(?:other)$/) {
1877 # set max outputs so linux can have up to 4 qxl displays with one device
1878 if (min_version
($machine_version, 4, 1)) {
1879 $max_outputs = ",max_outputs=4";
1884 die "no devicetype for $vga->{type}\n" if !$type;
1888 if ($vga->{type
} =~ /^virtio/) {
1889 my $bytes = PVE
::Tools
::convert_size
($vgamem_mb, "mb" => "b");
1890 $memory = ",max_hostmem=$bytes";
1892 # from https://www.spice-space.org/multiple-monitors.html
1893 $memory = ",vgamem_mb=$vga->{memory}";
1894 my $ram = $vgamem_mb * 4;
1895 my $vram = $vgamem_mb * 2;
1896 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
1898 $memory = ",vgamem_mb=$vga->{memory}";
1900 } elsif ($qxlnum && $id) {
1901 $memory = ",ram_size=67108864,vram_size=33554432";
1905 if ($type eq 'VGA' && windows_version
($conf->{ostype
})) {
1906 $edidoff=",edid=off" if (!defined($conf->{bios
}) || $conf->{bios
} ne 'ovmf');
1909 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
1910 my $vgaid = "vga" . ($id // '');
1912 if ($q35 && $vgaid eq 'vga') {
1913 # the first display uses pcie.0 bus on q35 machines
1914 $pciaddr = print_pcie_addr
($vgaid, $bridges, $arch, $machine);
1916 $pciaddr = print_pci_addr
($vgaid, $bridges, $arch, $machine);
1919 if ($vga->{type
} eq 'virtio-gl') {
1920 my $base = '/usr/lib/x86_64-linux-gnu/lib';
1921 die "missing libraries for '$vga->{type}' detected! Please install 'libgl1' and 'libegl1'\n"
1922 if !-e
"${base}EGL.so.1" || !-e
"${base}GL.so.1";
1924 die "no DRM render node detected (/dev/dri/renderD*), no GPU? - needed for '$vga->{type}' display\n"
1925 if !PVE
::Tools
::dir_glob_regex
('/dev/dri/', "renderD.*");
1928 return "$type,id=${vgaid}${memory}${max_outputs}${pciaddr}${edidoff}";
1931 sub parse_number_sets
{
1934 foreach my $part (split(/;/, $set)) {
1935 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1936 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1937 push @$res, [ $1, $2 ];
1939 die "invalid range: $part\n";
1948 my $res = parse_property_string
($numa_fmt, $data);
1949 $res->{cpus
} = parse_number_sets
($res->{cpus
}) if defined($res->{cpus
});
1950 $res->{hostnodes
} = parse_number_sets
($res->{hostnodes
}) if defined($res->{hostnodes
});
1954 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1956 my ($data, $disable_mac_autogen) = @_;
1958 my $res = eval { parse_property_string
($net_fmt, $data) };
1963 if (!defined($res->{macaddr
}) && !$disable_mac_autogen) {
1964 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
1965 $res->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
});
1970 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
1971 sub parse_ipconfig
{
1974 my $res = eval { parse_property_string
($ipconfig_fmt, $data) };
1980 if ($res->{gw
} && !$res->{ip
}) {
1981 warn 'gateway specified without specifying an IP address';
1984 if ($res->{gw6
} && !$res->{ip6
}) {
1985 warn 'IPv6 gateway specified without specifying an IPv6 address';
1988 if ($res->{gw
} && $res->{ip
} eq 'dhcp') {
1989 warn 'gateway specified together with DHCP';
1992 if ($res->{gw6
} && $res->{ip6
} !~ /^$IPV6RE/) {
1994 warn "IPv6 gateway specified together with $res->{ip6} address";
1998 if (!$res->{ip
} && !$res->{ip6
}) {
1999 return { ip
=> 'dhcp', ip6
=> 'dhcp' };
2008 return PVE
::JSONSchema
::print_property_string
($net, $net_fmt);
2011 sub add_random_macs
{
2012 my ($settings) = @_;
2014 foreach my $opt (keys %$settings) {
2015 next if $opt !~ m/^net(\d+)$/;
2016 my $net = parse_net
($settings->{$opt});
2018 $settings->{$opt} = print_net
($net);
2022 sub vm_is_volid_owner
{
2023 my ($storecfg, $vmid, $volid) = @_;
2025 if ($volid !~ m
|^/|) {
2027 eval { ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid); };
2028 if ($owner && ($owner == $vmid)) {
2036 sub vmconfig_register_unused_drive
{
2037 my ($storecfg, $vmid, $conf, $drive) = @_;
2039 if (drive_is_cloudinit
($drive)) {
2040 eval { PVE
::Storage
::vdisk_free
($storecfg, $drive->{file
}) };
2042 delete $conf->{cloudinit
};
2043 } elsif (!drive_is_cdrom
($drive)) {
2044 my $volid = $drive->{file
};
2045 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
2046 PVE
::QemuConfig-
>add_unused_volume($conf, $volid, $vmid);
2051 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2055 pattern
=> '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2056 format_description
=> 'UUID',
2057 description
=> "Set SMBIOS1 UUID.",
2062 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2063 format_description
=> 'Base64 encoded string',
2064 description
=> "Set SMBIOS1 version.",
2069 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2070 format_description
=> 'Base64 encoded string',
2071 description
=> "Set SMBIOS1 serial number.",
2076 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2077 format_description
=> 'Base64 encoded string',
2078 description
=> "Set SMBIOS1 manufacturer.",
2083 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2084 format_description
=> 'Base64 encoded string',
2085 description
=> "Set SMBIOS1 product ID.",
2090 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2091 format_description
=> 'Base64 encoded string',
2092 description
=> "Set SMBIOS1 SKU string.",
2097 pattern
=> '[A-Za-z0-9+\/]+={0,2}',
2098 format_description
=> 'Base64 encoded string',
2099 description
=> "Set SMBIOS1 family string.",
2104 description
=> 'Flag to indicate that the SMBIOS values are base64 encoded',
2112 my $res = eval { parse_property_string
($smbios1_fmt, $data) };
2119 return PVE
::JSONSchema
::print_property_string
($smbios1, $smbios1_fmt);
2122 PVE
::JSONSchema
::register_format
('pve-qm-smbios1', $smbios1_fmt);
2124 sub parse_watchdog
{
2129 my $res = eval { parse_property_string
($watchdog_fmt, $value) };
2134 sub parse_guest_agent
{
2137 return {} if !defined($conf->{agent
});
2139 my $res = eval { parse_property_string
($agent_fmt, $conf->{agent
}) };
2142 # if the agent is disabled ignore the other potentially set properties
2143 return {} if !$res->{enabled
};
2148 my ($conf, $key) = @_;
2149 return undef if !defined($conf->{agent
});
2151 my $agent = parse_guest_agent
($conf);
2152 return $agent->{$key};
2158 return {} if !$value;
2159 my $res = eval { parse_property_string
($vga_fmt, $value) };
2169 my $res = eval { parse_property_string
($rng_fmt, $value) };
2174 sub parse_meta_info
{
2179 my $res = eval { parse_property_string
($meta_info_fmt, $value) };
2184 sub new_meta_info_string
{
2185 my () = @_; # for now do not allow to override any value
2187 return PVE
::JSONSchema
::print_property_string
(
2189 'creation-qemu' => kvm_user_version
(),
2190 ctime
=> "". int(time()),
2196 sub qemu_created_version_fixups
{
2197 my ($conf, $forcemachine, $kvmver) = @_;
2199 my $meta = parse_meta_info
($conf->{meta
}) // {};
2200 my $forced_vers = PVE
::QemuServer
::Machine
::extract_version
($forcemachine);
2202 # check if we need to apply some handling for VMs that always use the latest machine version but
2203 # had a machine version transition happen that affected HW such that, e.g., an OS config change
2204 # would be required (we do not want to pin machine version for non-windows OS type)
2206 (!defined($conf->{machine
}) || $conf->{machine
} =~ m/^(?:pc|q35|virt)$/) # non-versioned machine
2207 && (!defined($meta->{'creation-qemu'}) || !min_version
($meta->{'creation-qemu'}, 6, 1)) # created before 6.1
2208 && (!$forced_vers || min_version
($forced_vers, 6, 1)) # handle snapshot-rollback/migrations
2209 && min_version
($kvmver, 6, 1) # only need to apply the change since 6.1
2211 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
2212 if ($q35 && $conf->{ostype
} && $conf->{ostype
} eq 'l26') {
2213 # this changed to default-on in Q 6.1 for q35 machines, it will mess with PCI slot view
2214 # and thus with the predictable interface naming of systemd
2215 return ['-global', 'ICH9-LPC.acpi-pci-hotplug-with-bridge-support=off'];
2221 PVE
::JSONSchema
::register_format
('pve-qm-usb-device', \
&verify_usb_device
);
2222 sub verify_usb_device
{
2223 my ($value, $noerr) = @_;
2225 return $value if parse_usb_device
($value);
2229 die "unable to parse usb device\n";
2232 # add JSON properties for create and set function
2233 sub json_config_properties
{
2234 my ($prop, $with_disk_alloc) = @_;
2236 my $skip_json_config_opts = {
2240 runningmachine
=> 1,
2245 foreach my $opt (keys %$confdesc) {
2246 next if $skip_json_config_opts->{$opt};
2248 if ($with_disk_alloc && is_valid_drivename
($opt)) {
2249 $prop->{$opt} = $PVE::QemuServer
::Drive
::drivedesc_hash_with_alloc-
>{$opt};
2251 $prop->{$opt} = $confdesc->{$opt};
2258 # Properties that we can read from an OVF file
2259 sub json_ovf_properties
{
2262 for my $device (PVE
::QemuServer
::Drive
::valid_drive_names
()) {
2263 $prop->{$device} = {
2265 format
=> 'pve-volume-id-or-absolute-path',
2266 description
=> "Disk image that gets imported to $device",
2273 description
=> "The number of CPU cores.",
2278 description
=> "Amount of RAM for the VM in MB.",
2283 description
=> "Name of the VM.",
2290 # return copy of $confdesc_cloudinit to generate documentation
2291 sub cloudinit_config_properties
{
2293 return dclone
($confdesc_cloudinit);
2296 sub cloudinit_pending_properties
{
2298 map { $_ => 1 } keys $confdesc_cloudinit->%*,
2301 $p->{"net$_"} = 1 for 0..($MAX_NETS-1);
2306 my ($key, $value) = @_;
2308 die "unknown setting '$key'\n" if !$confdesc->{$key};
2310 my $type = $confdesc->{$key}->{type
};
2312 if (!defined($value)) {
2313 die "got undefined value\n";
2316 if ($value =~ m/[\n\r]/) {
2317 die "property contains a line feed\n";
2320 if ($type eq 'boolean') {
2321 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2322 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2323 die "type check ('boolean') failed - got '$value'\n";
2324 } elsif ($type eq 'integer') {
2325 return int($1) if $value =~ m/^(\d+)$/;
2326 die "type check ('integer') failed - got '$value'\n";
2327 } elsif ($type eq 'number') {
2328 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2329 die "type check ('number') failed - got '$value'\n";
2330 } elsif ($type eq 'string') {
2331 if (my $fmt = $confdesc->{$key}->{format
}) {
2332 PVE
::JSONSchema
::check_format
($fmt, $value);
2335 $value =~ s/^\"(.*)\"$/$1/;
2338 die "internal error"
2343 my ($storecfg, $vmid, $skiplock, $replacement_conf, $purge_unreferenced) = @_;
2345 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2347 if (!$skiplock && !PVE
::QemuConfig-
>has_lock($conf, 'suspended')) {
2348 PVE
::QemuConfig-
>check_lock($conf);
2351 if ($conf->{template
}) {
2352 # check if any base image is still used by a linked clone
2353 PVE
::QemuConfig-
>foreach_volume_full($conf, { include_unused
=> 1 }, sub {
2354 my ($ds, $drive) = @_;
2355 return if drive_is_cdrom
($drive);
2357 my $volid = $drive->{file
};
2358 return if !$volid || $volid =~ m
|^/|;
2360 die "base volume '$volid' is still in use by linked cloned\n"
2361 if PVE
::Storage
::volume_is_base_and_used
($storecfg, $volid);
2367 my $remove_owned_drive = sub {
2368 my ($ds, $drive) = @_;
2369 return if drive_is_cdrom
($drive, 1);
2371 my $volid = $drive->{file
};
2372 return if !$volid || $volid =~ m
|^/|;
2373 return if $volids->{$volid};
2375 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
2376 return if !$path || !$owner || ($owner != $vmid);
2378 $volids->{$volid} = 1;
2379 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2380 warn "Could not remove disk '$volid', check manually: $@" if $@;
2383 # only remove disks owned by this VM (referenced in the config)
2384 my $include_opts = {
2385 include_unused
=> 1,
2386 extra_keys
=> ['vmstate'],
2388 PVE
::QemuConfig-
>foreach_volume_full($conf, $include_opts, $remove_owned_drive);
2390 for my $snap (values %{$conf->{snapshots
}}) {
2391 next if !defined($snap->{vmstate
});
2392 my $drive = PVE
::QemuConfig-
>parse_volume('vmstate', $snap->{vmstate
}, 1);
2393 next if !defined($drive);
2394 $remove_owned_drive->('vmstate', $drive);
2397 PVE
::QemuConfig-
>foreach_volume_full($conf->{pending
}, $include_opts, $remove_owned_drive);
2399 if ($purge_unreferenced) { # also remove unreferenced disk
2400 my $vmdisks = PVE
::Storage
::vdisk_list
($storecfg, undef, $vmid, undef, 'images');
2401 PVE
::Storage
::foreach_volid
($vmdisks, sub {
2402 my ($volid, $sid, $volname, $d) = @_;
2403 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid) };
2408 if (defined $replacement_conf) {
2409 PVE
::QemuConfig-
>write_config($vmid, $replacement_conf);
2411 PVE
::QemuConfig-
>destroy_config($vmid);
2415 sub parse_vm_config
{
2416 my ($filename, $raw, $strict) = @_;
2418 return if !defined($raw);
2421 digest
=> Digest
::SHA
::sha1_hex
($raw),
2427 my $handle_error = sub {
2437 $filename =~ m
|/qemu-server/(\d
+)\
.conf
$|
2438 || die "got strange filename '$filename'";
2444 my $finish_description = sub {
2445 if (defined($descr)) {
2447 $conf->{description
} = $descr;
2453 my @lines = split(/\n/, $raw);
2454 foreach my $line (@lines) {
2455 next if $line =~ m/^\s*$/;
2457 if ($line =~ m/^\[PENDING\]\s*$/i) {
2458 $section = 'pending';
2459 $finish_description->();
2460 $conf = $res->{$section} = {};
2462 } elsif ($line =~ m/^\[special:cloudinit\]\s*$/i) {
2463 $section = 'cloudinit';
2464 $finish_description->();
2465 $conf = $res->{$section} = {};
2468 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2470 $finish_description->();
2471 $conf = $res->{snapshots
}->{$section} = {};
2475 if ($line =~ m/^\#(.*)$/) {
2476 $descr = '' if !defined($descr);
2477 $descr .= PVE
::Tools
::decode_text
($1) . "\n";
2481 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2482 $descr = '' if !defined($descr);
2483 $descr .= PVE
::Tools
::decode_text
($2);
2484 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2485 $conf->{snapstate
} = $1;
2486 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2489 $conf->{$key} = $value;
2490 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2492 if ($section eq 'pending') {
2493 $conf->{delete} = $value; # we parse this later
2495 $handle_error->("vm $vmid - property 'delete' is only allowed in [PENDING]\n");
2497 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2500 if ($section eq 'cloudinit') {
2501 # ignore validation only used for informative purpose
2502 $conf->{$key} = $value;
2505 eval { $value = check_type
($key, $value); };
2507 $handle_error->("vm $vmid - unable to parse value of '$key' - $@");
2509 $key = 'ide2' if $key eq 'cdrom';
2510 my $fmt = $confdesc->{$key}->{format
};
2511 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2512 my $v = parse_drive
($key, $value);
2513 if (my $volid = filename_to_volume_id
($vmid, $v->{file
}, $v->{media
})) {
2514 $v->{file
} = $volid;
2515 $value = print_drive
($v);
2517 $handle_error->("vm $vmid - unable to parse value of '$key'\n");
2522 $conf->{$key} = $value;
2525 $handle_error->("vm $vmid - unable to parse config: $line\n");
2529 $finish_description->();
2530 delete $res->{snapstate
}; # just to be sure
2535 sub write_vm_config
{
2536 my ($filename, $conf) = @_;
2538 delete $conf->{snapstate
}; # just to be sure
2540 if ($conf->{cdrom
}) {
2541 die "option ide2 conflicts with cdrom\n" if $conf->{ide2
};
2542 $conf->{ide2
} = $conf->{cdrom
};
2543 delete $conf->{cdrom
};
2546 # we do not use 'smp' any longer
2547 if ($conf->{sockets
}) {
2548 delete $conf->{smp
};
2549 } elsif ($conf->{smp
}) {
2550 $conf->{sockets
} = $conf->{smp
};
2551 delete $conf->{cores
};
2552 delete $conf->{smp
};
2555 my $used_volids = {};
2557 my $cleanup_config = sub {
2558 my ($cref, $pending, $snapname) = @_;
2560 foreach my $key (keys %$cref) {
2561 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2562 $key eq 'snapstate' || $key eq 'pending' || $key eq 'cloudinit';
2563 my $value = $cref->{$key};
2564 if ($key eq 'delete') {
2565 die "propertry 'delete' is only allowed in [PENDING]\n"
2567 # fixme: check syntax?
2570 eval { $value = check_type
($key, $value); };
2571 die "unable to parse value of '$key' - $@" if $@;
2573 $cref->{$key} = $value;
2575 if (!$snapname && is_valid_drivename
($key)) {
2576 my $drive = parse_drive
($key, $value);
2577 $used_volids->{$drive->{file
}} = 1 if $drive && $drive->{file
};
2582 &$cleanup_config($conf);
2584 &$cleanup_config($conf->{pending
}, 1);
2586 foreach my $snapname (keys %{$conf->{snapshots
}}) {
2587 die "internal error: snapshot name '$snapname' is forbidden" if lc($snapname) eq 'pending';
2588 &$cleanup_config($conf->{snapshots
}->{$snapname}, undef, $snapname);
2591 # remove 'unusedX' settings if we re-add a volume
2592 foreach my $key (keys %$conf) {
2593 my $value = $conf->{$key};
2594 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2595 delete $conf->{$key};
2599 my $generate_raw_config = sub {
2600 my ($conf, $pending) = @_;
2604 # add description as comment to top of file
2605 if (defined(my $descr = $conf->{description
})) {
2607 foreach my $cl (split(/\n/, $descr)) {
2608 $raw .= '#' . PVE
::Tools
::encode_text
($cl) . "\n";
2611 $raw .= "#\n" if $pending;
2615 foreach my $key (sort keys %$conf) {
2616 next if $key =~ /^(digest|description|pending|cloudinit|snapshots)$/;
2617 $raw .= "$key: $conf->{$key}\n";
2622 my $raw = &$generate_raw_config($conf);
2624 if (scalar(keys %{$conf->{pending
}})){
2625 $raw .= "\n[PENDING]\n";
2626 $raw .= &$generate_raw_config($conf->{pending
}, 1);
2629 if (scalar(keys %{$conf->{cloudinit
}}) && PVE
::QemuConfig-
>has_cloudinit($conf)){
2630 $raw .= "\n[special:cloudinit]\n";
2631 $raw .= &$generate_raw_config($conf->{cloudinit
});
2634 foreach my $snapname (sort keys %{$conf->{snapshots
}}) {
2635 $raw .= "\n[$snapname]\n";
2636 $raw .= &$generate_raw_config($conf->{snapshots
}->{$snapname});
2646 # we use static defaults from our JSON schema configuration
2647 foreach my $key (keys %$confdesc) {
2648 if (defined(my $default = $confdesc->{$key}->{default})) {
2649 $res->{$key} = $default;
2657 my $vmlist = PVE
::Cluster
::get_vmlist
();
2659 return $res if !$vmlist || !$vmlist->{ids
};
2660 my $ids = $vmlist->{ids
};
2661 my $nodename = nodename
();
2663 foreach my $vmid (keys %$ids) {
2664 my $d = $ids->{$vmid};
2665 next if !$d->{node
} || $d->{node
} ne $nodename;
2666 next if !$d->{type
} || $d->{type
} ne 'qemu';
2667 $res->{$vmid}->{exists} = 1;
2672 # test if VM uses local resources (to prevent migration)
2673 sub check_local_resources
{
2674 my ($conf, $noerr) = @_;
2678 push @loc_res, "hostusb" if $conf->{hostusb
}; # old syntax
2679 push @loc_res, "hostpci" if $conf->{hostpci
}; # old syntax
2681 push @loc_res, "ivshmem" if $conf->{ivshmem
};
2683 foreach my $k (keys %$conf) {
2684 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2685 # sockets are safe: they will recreated be on the target side post-migrate
2686 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2687 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2690 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2695 # check if used storages are available on all nodes (use by migrate)
2696 sub check_storage_availability
{
2697 my ($storecfg, $conf, $node) = @_;
2699 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2700 my ($ds, $drive) = @_;
2702 my $volid = $drive->{file
};
2705 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2708 # check if storage is available on both nodes
2709 my $scfg = PVE
::Storage
::storage_check_enabled
($storecfg, $sid);
2710 PVE
::Storage
::storage_check_enabled
($storecfg, $sid, $node);
2712 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $volid);
2714 die "$volid: content type '$vtype' is not available on storage '$sid'\n"
2715 if !$scfg->{content
}->{$vtype};
2719 # list nodes where all VM images are available (used by has_feature API)
2721 my ($conf, $storecfg) = @_;
2723 my $nodelist = PVE
::Cluster
::get_nodelist
();
2724 my $nodehash = { map { $_ => 1 } @$nodelist };
2725 my $nodename = nodename
();
2727 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2728 my ($ds, $drive) = @_;
2730 my $volid = $drive->{file
};
2733 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2735 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2736 if ($scfg->{disable
}) {
2738 } elsif (my $avail = $scfg->{nodes
}) {
2739 foreach my $node (keys %$nodehash) {
2740 delete $nodehash->{$node} if !$avail->{$node};
2742 } elsif (!$scfg->{shared
}) {
2743 foreach my $node (keys %$nodehash) {
2744 delete $nodehash->{$node} if $node ne $nodename
2753 sub check_local_storage_availability
{
2754 my ($conf, $storecfg) = @_;
2756 my $nodelist = PVE
::Cluster
::get_nodelist
();
2757 my $nodehash = { map { $_ => {} } @$nodelist };
2759 PVE
::QemuConfig-
>foreach_volume($conf, sub {
2760 my ($ds, $drive) = @_;
2762 my $volid = $drive->{file
};
2765 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
2767 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
2769 if ($scfg->{disable
}) {
2770 foreach my $node (keys %$nodehash) {
2771 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2773 } elsif (my $avail = $scfg->{nodes
}) {
2774 foreach my $node (keys %$nodehash) {
2775 if (!$avail->{$node}) {
2776 $nodehash->{$node}->{unavailable_storages
}->{$storeid} = 1;
2783 foreach my $node (values %$nodehash) {
2784 if (my $unavail = $node->{unavailable_storages
}) {
2785 $node->{unavailable_storages
} = [ sort keys %$unavail ];
2792 # Compat only, use assert_config_exists_on_node and vm_running_locally where possible
2794 my ($vmid, $nocheck, $node) = @_;
2796 # $nocheck is set when called during a migration, in which case the config
2797 # file might still or already reside on the *other* node
2798 # - because rename has already happened, and current node is source
2799 # - because rename hasn't happened yet, and current node is target
2800 # - because rename has happened, current node is target, but hasn't yet
2802 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid, $node) if !$nocheck;
2803 return PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
2808 my $vzlist = config_list
();
2810 my $fd = IO
::Dir-
>new($PVE::QemuServer
::Helpers
::var_run_tmpdir
) || return $vzlist;
2812 while (defined(my $de = $fd->read)) {
2813 next if $de !~ m/^(\d+)\.pid$/;
2815 next if !defined($vzlist->{$vmid});
2816 if (my $pid = check_running
($vmid)) {
2817 $vzlist->{$vmid}->{pid
} = $pid;
2824 our $vmstatus_return_properties = {
2825 vmid
=> get_standard_option
('pve-vmid'),
2827 description
=> "QEMU process status.",
2829 enum
=> ['stopped', 'running'],
2832 description
=> "Maximum memory in bytes.",
2835 renderer
=> 'bytes',
2838 description
=> "Root disk size in bytes.",
2841 renderer
=> 'bytes',
2844 description
=> "VM name.",
2849 description
=> "QEMU QMP agent status.",
2854 description
=> "PID of running qemu process.",
2859 description
=> "Uptime.",
2862 renderer
=> 'duration',
2865 description
=> "Maximum usable CPUs.",
2870 description
=> "The current config lock, if any.",
2875 description
=> "The current configured tags, if any",
2879 'running-machine' => {
2880 description
=> "The currently running machine type (if running).",
2885 description
=> "The currently running QEMU version (if running).",
2891 my $last_proc_pid_stat;
2893 # get VM status information
2894 # This must be fast and should not block ($full == false)
2895 # We only query KVM using QMP if $full == true (this can be slow)
2897 my ($opt_vmid, $full) = @_;
2901 my $storecfg = PVE
::Storage
::config
();
2903 my $list = vzlist
();
2904 my $defaults = load_defaults
();
2906 my ($uptime) = PVE
::ProcFSTools
::read_proc_uptime
(1);
2908 my $cpucount = $cpuinfo->{cpus
} || 1;
2910 foreach my $vmid (keys %$list) {
2911 next if $opt_vmid && ($vmid ne $opt_vmid);
2913 my $conf = PVE
::QemuConfig-
>load_config($vmid);
2915 my $d = { vmid
=> int($vmid) };
2916 $d->{pid
} = int($list->{$vmid}->{pid
}) if $list->{$vmid}->{pid
};
2918 # fixme: better status?
2919 $d->{status
} = $list->{$vmid}->{pid
} ?
'running' : 'stopped';
2921 my $size = PVE
::QemuServer
::Drive
::bootdisk_size
($storecfg, $conf);
2922 if (defined($size)) {
2923 $d->{disk
} = 0; # no info available
2924 $d->{maxdisk
} = $size;
2930 $d->{cpus
} = ($conf->{sockets
} || $defaults->{sockets
})
2931 * ($conf->{cores
} || $defaults->{cores
});
2932 $d->{cpus
} = $cpucount if $d->{cpus
} > $cpucount;
2933 $d->{cpus
} = $conf->{vcpus
} if $conf->{vcpus
};
2935 $d->{name
} = $conf->{name
} || "VM $vmid";
2936 $d->{maxmem
} = $conf->{memory
} ?
$conf->{memory
}*(1024*1024)
2937 : $defaults->{memory
}*(1024*1024);
2939 if ($conf->{balloon
}) {
2940 $d->{balloon_min
} = $conf->{balloon
}*(1024*1024);
2941 $d->{shares
} = defined($conf->{shares
}) ?
$conf->{shares
}
2942 : $defaults->{shares
};
2953 $d->{diskwrite
} = 0;
2955 $d->{template
} = 1 if PVE
::QemuConfig-
>is_template($conf);
2957 $d->{serial
} = 1 if conf_has_serial
($conf);
2958 $d->{lock} = $conf->{lock} if $conf->{lock};
2959 $d->{tags
} = $conf->{tags
} if defined($conf->{tags
});
2964 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
2965 foreach my $dev (keys %$netdev) {
2966 next if $dev !~ m/^tap([1-9]\d*)i/;
2968 my $d = $res->{$vmid};
2971 $d->{netout
} += $netdev->{$dev}->{receive
};
2972 $d->{netin
} += $netdev->{$dev}->{transmit
};
2975 $d->{nics
}->{$dev}->{netout
} = int($netdev->{$dev}->{receive
});
2976 $d->{nics
}->{$dev}->{netin
} = int($netdev->{$dev}->{transmit
});
2981 my $ctime = gettimeofday
;
2983 foreach my $vmid (keys %$list) {
2985 my $d = $res->{$vmid};
2986 my $pid = $d->{pid
};
2989 my $pstat = PVE
::ProcFSTools
::read_proc_pid_stat
($pid);
2990 next if !$pstat; # not running
2992 my $used = $pstat->{utime} + $pstat->{stime
};
2994 $d->{uptime
} = int(($uptime - $pstat->{starttime
})/$cpuinfo->{user_hz
});
2996 if ($pstat->{vsize
}) {
2997 $d->{mem
} = int(($pstat->{rss
}/$pstat->{vsize
})*$d->{maxmem
});
3000 my $old = $last_proc_pid_stat->{$pid};
3002 $last_proc_pid_stat->{$pid} = {
3010 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
3012 if ($dtime > 1000) {
3013 my $dutime = $used - $old->{used
};
3015 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
3016 $last_proc_pid_stat->{$pid} = {
3022 $d->{cpu
} = $old->{cpu
};
3026 return $res if !$full;
3028 my $qmpclient = PVE
::QMPClient-
>new();
3030 my $ballooncb = sub {
3031 my ($vmid, $resp) = @_;
3033 my $info = $resp->{'return'};
3034 return if !$info->{max_mem
};
3036 my $d = $res->{$vmid};
3038 # use memory assigned to VM
3039 $d->{maxmem
} = $info->{max_mem
};
3040 $d->{balloon
} = $info->{actual
};
3042 if (defined($info->{total_mem
}) && defined($info->{free_mem
})) {
3043 $d->{mem
} = $info->{total_mem
} - $info->{free_mem
};
3044 $d->{freemem
} = $info->{free_mem
};
3047 $d->{ballooninfo
} = $info;
3050 my $blockstatscb = sub {
3051 my ($vmid, $resp) = @_;
3052 my $data = $resp->{'return'} || [];
3053 my $totalrdbytes = 0;
3054 my $totalwrbytes = 0;
3056 for my $blockstat (@$data) {
3057 $totalrdbytes = $totalrdbytes + $blockstat->{stats
}->{rd_bytes
};
3058 $totalwrbytes = $totalwrbytes + $blockstat->{stats
}->{wr_bytes
};
3060 $blockstat->{device
} =~ s/drive-//;
3061 $res->{$vmid}->{blockstat
}->{$blockstat->{device
}} = $blockstat->{stats
};
3063 $res->{$vmid}->{diskread
} = $totalrdbytes;
3064 $res->{$vmid}->{diskwrite
} = $totalwrbytes;
3067 my $machinecb = sub {
3068 my ($vmid, $resp) = @_;
3069 my $data = $resp->{'return'} || [];
3071 $res->{$vmid}->{'running-machine'} =
3072 PVE
::QemuServer
::Machine
::current_from_query_machines
($data);
3075 my $versioncb = sub {
3076 my ($vmid, $resp) = @_;
3077 my $data = $resp->{'return'} // {};
3078 my $version = 'unknown';
3080 if (my $v = $data->{qemu
}) {
3081 $version = $v->{major
} . "." . $v->{minor
} . "." . $v->{micro
};
3084 $res->{$vmid}->{'running-qemu'} = $version;
3087 my $statuscb = sub {
3088 my ($vmid, $resp) = @_;
3090 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3091 $qmpclient->queue_cmd($vmid, $machinecb, 'query-machines');
3092 $qmpclient->queue_cmd($vmid, $versioncb, 'query-version');
3093 # this fails if ballon driver is not loaded, so this must be
3094 # the last commnand (following command are aborted if this fails).
3095 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3097 my $status = 'unknown';
3098 if (!defined($status = $resp->{'return'}->{status
})) {
3099 warn "unable to get VM status\n";
3103 $res->{$vmid}->{qmpstatus
} = $resp->{'return'}->{status
};
3106 foreach my $vmid (keys %$list) {
3107 next if $opt_vmid && ($vmid ne $opt_vmid);
3108 next if !$res->{$vmid}->{pid
}; # not running
3109 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3112 $qmpclient->queue_execute(undef, 2);
3114 foreach my $vmid (keys %$list) {
3115 next if $opt_vmid && ($vmid ne $opt_vmid);
3116 next if !$res->{$vmid}->{pid
}; #not running
3118 # we can't use the $qmpclient since it might have already aborted on
3119 # 'query-balloon', but this might also fail for older versions...
3120 my $qemu_support = eval { mon_cmd
($vmid, "query-proxmox-support") };
3121 $res->{$vmid}->{'proxmox-support'} = $qemu_support // {};
3124 foreach my $vmid (keys %$list) {
3125 next if $opt_vmid && ($vmid ne $opt_vmid);
3126 $res->{$vmid}->{qmpstatus
} = $res->{$vmid}->{status
} if !$res->{$vmid}->{qmpstatus
};
3132 sub conf_has_serial
{
3135 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3136 if ($conf->{"serial$i"}) {
3144 sub conf_has_audio
{
3145 my ($conf, $id) = @_;
3148 my $audio = $conf->{"audio$id"};
3149 return if !defined($audio);
3151 my $audioproperties = parse_property_string
($audio_fmt, $audio);
3152 my $audiodriver = $audioproperties->{driver
} // 'spice';
3155 dev
=> $audioproperties->{device
},
3156 dev_id
=> "audiodev$id",
3157 backend
=> $audiodriver,
3158 backend_id
=> "$audiodriver-backend${id}",
3163 my ($audio, $audiopciaddr, $machine_version) = @_;
3167 my $id = $audio->{dev_id
};
3169 if (min_version
($machine_version, 4, 2)) {
3170 $audiodev = ",audiodev=$audio->{backend_id}";
3173 if ($audio->{dev
} eq 'AC97') {
3174 push @$devs, '-device', "AC97,id=${id}${audiopciaddr}$audiodev";
3175 } elsif ($audio->{dev
} =~ /intel\-hda$/) {
3176 push @$devs, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3177 push @$devs, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0$audiodev";
3178 push @$devs, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1$audiodev";
3180 die "unkown audio device '$audio->{dev}', implement me!";
3183 push @$devs, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3191 socket => "/var/run/qemu-server/$vmid.swtpm",
3192 pid
=> "/var/run/qemu-server/$vmid.swtpm.pid",
3196 sub add_tpm_device
{
3197 my ($vmid, $devices, $conf) = @_;
3199 return if !$conf->{tpmstate0
};
3201 my $paths = get_tpm_paths
($vmid);
3203 push @$devices, "-chardev", "socket,id=tpmchar,path=$paths->{socket}";
3204 push @$devices, "-tpmdev", "emulator,id=tpmdev,chardev=tpmchar";
3205 push @$devices, "-device", "tpm-tis,tpmdev=tpmdev";
3209 my ($storecfg, $vmid, $tpmdrive, $migration) = @_;
3211 return if !$tpmdrive;
3214 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
3215 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
3217 $state = PVE
::Storage
::map_volume
($storecfg, $tpm->{file
});
3219 $state = $tpm->{file
};
3222 my $paths = get_tpm_paths
($vmid);
3224 # during migration, we will get state from remote
3227 # run swtpm_setup to create a new TPM state if it doesn't exist yet
3234 "--create-platform-cert",
3237 "/etc/swtpm_setup.conf", # do not use XDG configs
3239 "0", # force creation as root, error if not possible
3240 "--not-overwrite", # ignore existing state, do not modify
3243 push @$setup_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3244 # TPM 2.0 supports ECC crypto, use if possible
3245 push @$setup_cmd, "--ecc" if $tpm->{version
} eq 'v2.0';
3247 run_command
($setup_cmd, outfunc
=> sub {
3248 print "swtpm_setup: $1\n";
3252 my $emulator_cmd = [
3256 "backend-uri=file://$state,mode=0600",
3258 "type=unixio,path=$paths->{socket},mode=0600",
3260 "file=$paths->{pid}",
3261 "--terminate", # terminate on QEMU disconnect
3264 push @$emulator_cmd, "--tpm2" if $tpm->{version
} eq 'v2.0';
3265 run_command
($emulator_cmd, outfunc
=> sub { print $1; });
3267 my $tries = 100; # swtpm may take a bit to start before daemonizing, wait up to 5s for pid
3268 while (! -e
$paths->{pid
}) {
3269 die "failed to start swtpm: pid file '$paths->{pid}' wasn't created.\n" if --$tries == 0;
3273 # return untainted PID of swtpm daemon so it can be killed on error
3274 file_read_firstline
($paths->{pid
}) =~ m/(\d+)/;
3278 sub vga_conf_has_spice
{
3281 my $vgaconf = parse_vga
($vga);
3282 my $vgatype = $vgaconf->{type
};
3283 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3290 return get_host_arch
() eq $arch;
3295 return $conf->{arch
} // get_host_arch
();
3298 my $default_machines = {
3303 sub get_installed_machine_version
{
3304 my ($kvmversion) = @_;
3305 $kvmversion = kvm_user_version
() if !defined($kvmversion);
3306 $kvmversion =~ m/^(\d+\.\d+)/;
3310 sub windows_get_pinned_machine_version
{
3311 my ($machine, $base_version, $kvmversion) = @_;
3313 my $pin_version = $base_version;
3314 if (!defined($base_version) ||
3315 !PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($base_version, $kvmversion)
3317 $pin_version = get_installed_machine_version
($kvmversion);
3319 if (!$machine || $machine eq 'pc') {
3320 $machine = "pc-i440fx-$pin_version";
3321 } elsif ($machine eq 'q35') {
3322 $machine = "pc-q35-$pin_version";
3323 } elsif ($machine eq 'virt') {
3324 $machine = "virt-$pin_version";
3326 warn "unknown machine type '$machine', not touching that!\n";
3332 sub get_vm_machine
{
3333 my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
3335 my $machine = $forcemachine || $conf->{machine
};
3337 if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
3338 $kvmversion //= kvm_user_version
();
3339 # we must pin Windows VMs without a specific version to 5.1, as 5.2 fixed a bug in ACPI
3340 # layout which confuses windows quite a bit and may result in various regressions..
3341 # see: https://lists.gnu.org/archive/html/qemu-devel/2021-02/msg08484.html
3342 if (windows_version
($conf->{ostype
})) {
3343 $machine = windows_get_pinned_machine_version
($machine, '5.1', $kvmversion);
3346 $machine ||= $default_machines->{$arch};
3347 if ($add_pve_version) {
3348 my $pvever = PVE
::QemuServer
::Machine
::get_pve_version
($kvmversion);
3349 $machine .= "+pve$pvever";
3353 if ($add_pve_version && $machine !~ m/\+pve\d+?(?:\.pxe)?$/) {
3354 my $is_pxe = $machine =~ m/^(.*?)\.pxe$/;
3355 $machine = $1 if $is_pxe;
3357 # for version-pinned machines that do not include a pve-version (e.g.
3358 # pc-q35-4.1), we assume 0 to keep them stable in case we bump
3359 $machine .= '+pve0';
3361 $machine .= '.pxe' if $is_pxe;
3367 sub get_ovmf_files
($$$) {
3368 my ($arch, $efidisk, $smm) = @_;
3370 my $types = $OVMF->{$arch}
3371 or die "no OVMF images known for architecture '$arch'\n";
3373 my $type = 'default';
3374 if (defined($efidisk->{efitype
}) && $efidisk->{efitype
} eq '4m') {
3375 $type = $smm ?
"4m" : "4m-no-smm";
3376 $type .= '-ms' if $efidisk->{'pre-enrolled-keys'};
3379 my ($ovmf_code, $ovmf_vars) = $types->{$type}->@*;
3380 die "EFI base image '$ovmf_code' not found\n" if ! -f
$ovmf_code;
3381 die "EFI vars image '$ovmf_vars' not found\n" if ! -f
$ovmf_vars;
3383 return ($ovmf_code, $ovmf_vars);
3387 aarch64
=> '/usr/bin/qemu-system-aarch64',
3388 x86_64
=> '/usr/bin/qemu-system-x86_64',
3390 sub get_command_for_arch
($) {
3392 return '/usr/bin/kvm' if is_native
($arch);
3394 my $cmd = $Arch2Qemu->{$arch}
3395 or die "don't know how to emulate architecture '$arch'\n";
3399 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
3400 # to use in a QEMU command line (-cpu element), first array_intersect the result
3401 # of query_supported_ with query_understood_. This is necessary because:
3403 # a) query_understood_ returns flags the host cannot use and
3404 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
3405 # flags, but CPU settings - with most of them being flags. Those settings
3406 # (and some flags, curiously) cannot be specified as a "-cpu" argument.
3408 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
3409 # expensive. If you need the value returned from this, you can get it much
3410 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
3411 # $accel being 'kvm' or 'tcg'.
3413 # pvestatd calls this function on startup and whenever the QEMU/KVM version
3414 # changes, automatically populating pmxcfs.
3416 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
3417 # since kvm and tcg machines support different flags
3419 sub query_supported_cpu_flags
{
3422 $arch //= get_host_arch
();
3423 my $default_machine = $default_machines->{$arch};
3427 # FIXME: Once this is merged, the code below should work for ARM as well:
3428 # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
3429 die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
3432 my $kvm_supported = defined(kvm_version
());
3433 my $qemu_cmd = get_command_for_arch
($arch);
3435 my $pidfile = PVE
::QemuServer
::Helpers
::pidfile_name
($fakevmid);
3437 # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
3438 my $query_supported_run_qemu = sub {
3444 '-machine', $default_machine,
3446 '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server=on,wait=off",
3447 '-mon', 'chardev=qmp,mode=control',
3448 '-pidfile', $pidfile,
3453 push @$cmd, '-accel', 'tcg';
3456 my $rc = run_command
($cmd, noerr
=> 1, quiet
=> 0);
3457 die "QEMU flag querying VM exited with code " . $rc if $rc;
3460 my $cmd_result = mon_cmd
(
3462 'query-cpu-model-expansion',
3464 model
=> { name
=> 'host' }
3467 my $props = $cmd_result->{model
}->{props
};
3468 foreach my $prop (keys %$props) {
3469 next if $props->{$prop} ne '1';
3470 # QEMU returns some flags multiple times, with '_', '.' or '-'
3471 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
3472 # We only keep those with underscores, to match /proc/cpuinfo
3473 $prop =~ s/\.|-/_/g;
3474 $flags->{$prop} = 1;
3479 # force stop with 10 sec timeout and 'nocheck', always stop, even if QMP failed
3480 vm_stop
(undef, $fakevmid, 1, 1, 10, 0, 1);
3484 return [ sort keys %$flags ];
3487 # We need to query QEMU twice, since KVM and TCG have different supported flags
3488 PVE
::QemuConfig-
>lock_config($fakevmid, sub {
3489 $flags->{tcg
} = eval { $query_supported_run_qemu->(0) };
3490 warn "warning: failed querying supported tcg flags: $@\n" if $@;
3492 if ($kvm_supported) {
3493 $flags->{kvm
} = eval { $query_supported_run_qemu->(1) };
3494 warn "warning: failed querying supported kvm flags: $@\n" if $@;
3501 # Understood CPU flags are written to a file at 'pve-qemu' compile time
3502 my $understood_cpu_flag_dir = "/usr/share/kvm";
3503 sub query_understood_cpu_flags
{
3504 my $arch = get_host_arch
();
3505 my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
3507 die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
3510 my $raw = file_get_contents
($filepath);
3511 $raw =~ s/^\s+|\s+$//g;
3512 my @flags = split(/\s+/, $raw);
3517 # Since commit 277d33454f77ec1d1e0bc04e37621e4dd2424b67 in pve-qemu, smm is not off by default
3518 # anymore. But smm=off seems to be required when using SeaBIOS and serial display.
3519 my sub should_disable_smm
{
3520 my ($conf, $vga) = @_;
3522 return (!defined($conf->{bios
}) || $conf->{bios
} eq 'seabios') &&
3523 $vga->{type
} && $vga->{type
} =~ m/^(serial\d+|none)$/;
3526 my sub print_ovmf_drive_commandlines
{
3527 my ($conf, $storecfg, $vmid, $arch, $q35, $version_guard) = @_;
3529 my $d = $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
3531 my ($ovmf_code, $ovmf_vars) = get_ovmf_files
($arch, $d, $q35);
3533 my $var_drive_str = "if=pflash,unit=1,id=drive-efidisk0";
3535 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($d->{file
}, 1);
3536 my ($path, $format) = $d->@{'file', 'format'};
3538 $path = PVE
::Storage
::path
($storecfg, $d->{file
});
3539 if (!defined($format)) {
3540 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
3541 $format = qemu_img_format
($scfg, $volname);
3543 } elsif (!defined($format)) {
3544 die "efidisk format must be specified\n";
3546 # SPI flash does lots of read-modify-write OPs, without writeback this gets really slow #3329
3547 if ($path =~ m/^rbd:/) {
3548 $var_drive_str .= ',cache=writeback';
3549 $path .= ':rbd_cache_policy=writeback'; # avoid write-around, we *need* to cache writes too
3551 $var_drive_str .= ",format=$format,file=$path";
3553 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $format eq 'raw' && $version_guard->(4, 1, 2);
3554 $var_drive_str .= ',readonly=on' if drive_is_read_only
($conf, $d);
3556 log_warn
("no efidisk configured! Using temporary efivars disk.");
3557 my $path = "/tmp/$vmid-ovmf.fd";
3558 PVE
::Tools
::file_copy
($ovmf_vars, $path, -s
$ovmf_vars);
3559 $var_drive_str .= ",format=raw,file=$path";
3560 $var_drive_str .= ",size=" . (-s
$ovmf_vars) if $version_guard->(4, 1, 2);
3563 return ("if=pflash,unit=0,format=raw,readonly=on,file=$ovmf_code", $var_drive_str);
3566 sub config_to_command
{
3567 my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu,
3570 my ($globalFlags, $machineFlags, $rtcFlags) = ([], [], []);
3573 my $ostype = $conf->{ostype
};
3574 my $winversion = windows_version
($ostype);
3575 my $kvm = $conf->{kvm
};
3576 my $nodename = nodename
();
3578 my $arch = get_vm_arch
($conf);
3579 my $kvm_binary = get_command_for_arch
($arch);
3580 my $kvmver = kvm_user_version
($kvm_binary);
3582 if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
3583 $kvmver //= "undefined";
3584 die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
3587 my $add_pve_version = min_version
($kvmver, 4, 1);
3589 my $machine_type = get_vm_machine
($conf, $forcemachine, $arch, $add_pve_version);
3590 my $machine_version = extract_version
($machine_type, $kvmver);
3591 $kvm //= 1 if is_native
($arch);
3593 $machine_version =~ m/(\d+)\.(\d+)/;
3594 my ($machine_major, $machine_minor) = ($1, $2);
3596 if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
3597 warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
3598 } elsif (!min_version
($kvmver, $machine_major, $machine_minor)) {
3599 die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type',"
3600 ." please upgrade node '$nodename'\n"
3601 } elsif (!PVE
::QemuServer
::Machine
::can_run_pve_machine_version
($machine_version, $kvmver)) {
3602 my $max_pve_version = PVE
::QemuServer
::Machine
::get_pve_version
($machine_version);
3603 die "Installed qemu-server (max feature level for $machine_major.$machine_minor is"
3604 ." pve$max_pve_version) is too old to run machine type '$machine_type', please upgrade"
3605 ." node '$nodename'\n";
3608 # if a specific +pve version is required for a feature, use $version_guard
3609 # instead of min_version to allow machines to be run with the minimum
3611 my $required_pve_version = 0;
3612 my $version_guard = sub {
3613 my ($major, $minor, $pve) = @_;
3614 return 0 if !min_version
($machine_version, $major, $minor, $pve);
3615 my $max_pve = PVE
::QemuServer
::Machine
::get_pve_version
("$major.$minor");
3616 return 1 if min_version
($machine_version, $major, $minor, $max_pve+1);
3617 $required_pve_version = $pve if $pve && $pve > $required_pve_version;
3621 if ($kvm && !defined kvm_version
()) {
3622 die "KVM virtualisation configured, but not available. Either disable in VM configuration"
3623 ." or enable in BIOS.\n";
3626 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
3627 my $hotplug_features = parse_hotplug_features
(defined($conf->{hotplug
}) ?
$conf->{hotplug
} : '1');
3628 my $use_old_bios_files = undef;
3629 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
3632 if ($conf->{affinity
}) {
3633 push @$cmd, '/usr/bin/taskset', '--cpu-list', '--all-tasks', $conf->{affinity
};
3636 push @$cmd, $kvm_binary;
3638 push @$cmd, '-id', $vmid;
3640 my $vmname = $conf->{name
} || "vm$vmid";
3642 push @$cmd, '-name', "$vmname,debug-threads=on";
3644 push @$cmd, '-no-shutdown';
3648 my $qmpsocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid);
3649 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server=on,wait=off";
3650 push @$cmd, '-mon', "chardev=qmp,mode=control";
3652 if (min_version
($machine_version, 2, 12)) {
3653 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3654 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3657 push @$cmd, '-pidfile' , PVE
::QemuServer
::Helpers
::pidfile_name
($vmid);
3659 push @$cmd, '-daemonize';
3661 if ($conf->{smbios1
}) {
3662 my $smbios_conf = parse_smbios1
($conf->{smbios1
});
3663 if ($smbios_conf->{base64
}) {
3664 # Do not pass base64 flag to qemu
3665 delete $smbios_conf->{base64
};
3666 my $smbios_string = "";
3667 foreach my $key (keys %$smbios_conf) {
3669 if ($key eq "uuid") {
3670 $value = $smbios_conf->{uuid
}
3672 $value = decode_base64
($smbios_conf->{$key});
3674 # qemu accepts any binary data, only commas need escaping by double comma
3676 $smbios_string .= "," . $key . "=" . $value if $value;
3678 push @$cmd, '-smbios', "type=1" . $smbios_string;
3680 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3684 if ($conf->{bios
} && $conf->{bios
} eq 'ovmf') {
3685 my ($code_drive_str, $var_drive_str) =
3686 print_ovmf_drive_commandlines
($conf, $storecfg, $vmid, $arch, $q35, $version_guard);
3687 push $cmd->@*, '-drive', $code_drive_str;
3688 push $cmd->@*, '-drive', $var_drive_str;
3691 if ($q35) { # tell QEMU to load q35 config early
3692 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3693 if (min_version
($machine_version, 4, 0)) {
3694 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3696 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3700 if (defined(my $fixups = qemu_created_version_fixups
($conf, $forcemachine, $kvmver))) {
3701 push @$cmd, $fixups->@*;
3704 if ($conf->{vmgenid
}) {
3705 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid
};
3708 # add usb controllers
3709 my @usbcontrollers = PVE
::QemuServer
::USB
::get_usb_controllers
(
3710 $conf, $bridges, $arch, $machine_type, $usbdesc->{format
}, $MAX_USB_DEVICES, $machine_version);
3711 push @$devices, @usbcontrollers if @usbcontrollers;
3712 my $vga = parse_vga
($conf->{vga
});
3714 my $qxlnum = vga_conf_has_spice
($conf->{vga
});
3715 $vga->{type
} = 'qxl' if $qxlnum;
3717 if (!$vga->{type
}) {
3718 if ($arch eq 'aarch64') {
3719 $vga->{type
} = 'virtio';
3720 } elsif (min_version
($machine_version, 2, 9)) {
3721 $vga->{type
} = (!$winversion || $winversion >= 6) ?
'std' : 'cirrus';
3723 $vga->{type
} = ($winversion >= 6) ?
'std' : 'cirrus';
3727 # enable absolute mouse coordinates (needed by vnc)
3728 my $tablet = $conf->{tablet
};
3729 if (!defined($tablet)) {
3730 $tablet = $defaults->{tablet
};
3731 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3732 $tablet = 0 if $vga->{type
} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3736 push @$devices, '-device', print_tabletdevice_full
($conf, $arch) if $tablet;
3737 my $kbd = print_keyboarddevice_full
($conf, $arch);
3738 push @$devices, '-device', $kbd if defined($kbd);
3741 my $bootorder = device_bootorder
($conf);
3743 # host pci device passthrough
3744 my ($kvm_off, $gpu_passthrough, $legacy_igd) = PVE
::QemuServer
::PCI
::print_hostpci_devices
(
3745 $vmid, $conf, $devices, $vga, $winversion, $q35, $bridges, $arch, $machine_type, $bootorder);
3748 my $usb_dev_features = {};
3749 $usb_dev_features->{spice_usb3
} = 1 if min_version
($machine_version, 4, 0);
3751 my @usbdevices = PVE
::QemuServer
::USB
::get_usb_devices
(
3752 $conf, $usbdesc->{format
}, $MAX_USB_DEVICES, $usb_dev_features, $bootorder, $machine_version);
3753 push @$devices, @usbdevices if @usbdevices;
3756 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3757 my $path = $conf->{"serial$i"} or next;
3758 if ($path eq 'socket') {
3759 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3760 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server=on,wait=off";
3761 # On aarch64, serial0 is the UART device. QEMU only allows
3762 # connecting UART devices via the '-serial' command line, as
3763 # the device has a fixed slot on the hardware...
3764 if ($arch eq 'aarch64' && $i == 0) {
3765 push @$devices, '-serial', "chardev:serial$i";
3767 push @$devices, '-device', "isa-serial,chardev=serial$i";
3770 die "no such serial device\n" if ! -c
$path;
3771 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3772 push @$devices, '-device', "isa-serial,chardev=serial$i";
3777 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3778 if (my $path = $conf->{"parallel$i"}) {
3779 die "no such parallel device\n" if ! -c
$path;
3780 my $devtype = $path =~ m!^/dev/usb/lp! ?
'tty' : 'parport';
3781 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3782 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3786 if (min_version
($machine_version, 4, 0) && (my $audio = conf_has_audio
($conf))) {
3787 my $audiopciaddr = print_pci_addr
("audio0", $bridges, $arch, $machine_type);
3788 my $audio_devs = audio_devs
($audio, $audiopciaddr, $machine_version);
3789 push @$devices, @$audio_devs;
3792 add_tpm_device
($vmid, $devices, $conf);
3795 $sockets = $conf->{smp
} if $conf->{smp
}; # old style - no longer iused
3796 $sockets = $conf->{sockets
} if $conf->{sockets
};
3798 my $cores = $conf->{cores
} || 1;
3800 my $maxcpus = $sockets * $cores;
3802 my $vcpus = $conf->{vcpus
} ?
$conf->{vcpus
} : $maxcpus;
3804 my $allowed_vcpus = $cpuinfo->{cpus
};
3806 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n" if ($allowed_vcpus < $maxcpus);
3808 if ($hotplug_features->{cpu
} && min_version
($machine_version, 2, 7)) {
3809 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3810 for (my $i = 2; $i <= $vcpus; $i++) {
3811 my $cpustr = print_cpu_device
($conf,$i);
3812 push @$cmd, '-device', $cpustr;
3817 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3819 push @$cmd, '-nodefaults';
3821 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3823 push @$cmd, '-no-acpi' if defined($conf->{acpi
}) && $conf->{acpi
} == 0;
3825 push @$cmd, '-no-reboot' if defined($conf->{reboot
}) && $conf->{reboot
} == 0;
3827 if ($vga->{type
} && $vga->{type
} !~ m/^serial\d+$/ && $vga->{type
} ne 'none'){
3828 push @$devices, '-device', print_vga_device
(
3829 $conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3831 push @$cmd, '-display', 'egl-headless,gl=core' if $vga->{type
} eq 'virtio-gl'; # VIRGL
3833 my $socket = PVE
::QemuServer
::Helpers
::vnc_socket
($vmid);
3834 push @$cmd, '-vnc', "unix:$socket,password=on";
3836 push @$cmd, '-vga', 'none' if $vga->{type
} eq 'none';
3837 push @$cmd, '-nographic';
3841 my $tdf = defined($conf->{tdf
}) ?
$conf->{tdf
} : $defaults->{tdf
};
3842 my $useLocaltime = $conf->{localtime};
3844 if ($winversion >= 5) { # windows
3845 $useLocaltime = 1 if !defined($conf->{localtime});
3847 # use time drift fix when acpi is enabled
3848 if (!(defined($conf->{acpi
}) && $conf->{acpi
} == 0)) {
3849 $tdf = 1 if !defined($conf->{tdf
});
3853 if ($winversion >= 6) {
3854 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3855 push @$cmd, '-no-hpet';
3858 push @$rtcFlags, 'driftfix=slew' if $tdf;
3860 if ($conf->{startdate
} && $conf->{startdate
} ne 'now') {
3861 push @$rtcFlags, "base=$conf->{startdate}";
3862 } elsif ($useLocaltime) {
3863 push @$rtcFlags, 'base=localtime';
3867 push @$cmd, '-cpu', $forcecpu;
3869 push @$cmd, get_cpu_options
($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough);
3872 PVE
::QemuServer
::Memory
::config
($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3874 push @$cmd, '-S' if $conf->{freeze
};
3876 push @$cmd, '-k', $conf->{keyboard
} if defined($conf->{keyboard
});
3878 my $guest_agent = parse_guest_agent
($conf);
3880 if ($guest_agent->{enabled
}) {
3881 my $qgasocket = PVE
::QemuServer
::Helpers
::qmp_socket
($vmid, 1);
3882 push @$devices, '-chardev', "socket,path=$qgasocket,server=on,wait=off,id=qga0";
3884 if (!$guest_agent->{type
} || $guest_agent->{type
} eq 'virtio') {
3885 my $pciaddr = print_pci_addr
("qga0", $bridges, $arch, $machine_type);
3886 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3887 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3888 } elsif ($guest_agent->{type
} eq 'isa') {
3889 push @$devices, '-device', "isa-serial,chardev=qga0";
3893 my $rng = $conf->{rng0
} ? parse_rng
($conf->{rng0
}) : undef;
3894 if ($rng && $version_guard->(4, 1, 2)) {
3895 check_rng_source
($rng->{source
});
3897 my $max_bytes = $rng->{max_bytes
} // $rng_fmt->{max_bytes
}->{default};
3898 my $period = $rng->{period
} // $rng_fmt->{period
}->{default};
3899 my $limiter_str = "";
3901 $limiter_str = ",max-bytes=$max_bytes,period=$period";
3904 my $rng_addr = print_pci_addr
("rng0", $bridges, $arch, $machine_type);
3905 push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3906 push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3911 if ($qxlnum || $vga->{type
} =~ /^virtio/) {
3914 for (my $i = 1; $i < $qxlnum; $i++){
3915 push @$devices, '-device', print_vga_device
(
3916 $conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3919 # assume other OS works like Linux
3920 my ($ram, $vram) = ("134217728", "67108864");
3921 if ($vga->{memory
}) {
3922 $ram = PVE
::Tools
::convert_size
($qxlnum*4*$vga->{memory
}, 'mb' => 'b');
3923 $vram = PVE
::Tools
::convert_size
($qxlnum*2*$vga->{memory
}, 'mb' => 'b');
3925 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3926 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3930 my $pciaddr = print_pci_addr
("spice", $bridges, $arch, $machine_type);
3932 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
3933 my @nodeaddrs = PVE
::Tools
::getaddrinfo_all
('localhost', family
=> $pfamily);
3934 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3936 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3937 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3938 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3940 my $localhost = PVE
::Network
::addr_to_ip
($nodeaddrs[0]->{addr
});
3941 $spice_port = PVE
::Tools
::next_spice_port
($pfamily, $localhost);
3943 my $spice_enhancement_str = $conf->{spice_enhancements
} // '';
3944 my $spice_enhancement = parse_property_string
($spice_enhancements_fmt, $spice_enhancement_str);
3945 if ($spice_enhancement->{foldersharing
}) {
3946 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3947 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3950 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3951 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}"
3952 if $spice_enhancement->{videostreaming
};
3954 push @$devices, '-spice', "$spice_opts";
3957 # enable balloon by default, unless explicitly disabled
3958 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
3959 my $pciaddr = print_pci_addr
("balloon0", $bridges, $arch, $machine_type);
3960 my $ballooncmd = "virtio-balloon-pci,id=balloon0$pciaddr";
3961 $ballooncmd .= ",free-page-reporting=on" if min_version
($machine_version, 6, 2);
3962 push @$devices, '-device', $ballooncmd;
3965 if ($conf->{watchdog
}) {
3966 my $wdopts = parse_watchdog
($conf->{watchdog
});
3967 my $pciaddr = print_pci_addr
("watchdog", $bridges, $arch, $machine_type);
3968 my $watchdog = $wdopts->{model
} || 'i6300esb';
3969 push @$devices, '-device', "$watchdog$pciaddr";
3970 push @$devices, '-watchdog-action', $wdopts->{action
} if $wdopts->{action
};
3974 my $scsicontroller = {};
3975 my $ahcicontroller = {};
3976 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : $defaults->{scsihw
};
3978 # Add iscsi initiator name if available
3979 if (my $initiator = get_initiator_name
()) {
3980 push @$devices, '-iscsi', "initiator-name=$initiator";
3983 PVE
::QemuConfig-
>foreach_volume($conf, sub {
3984 my ($ds, $drive) = @_;
3986 if (PVE
::Storage
::parse_volume_id
($drive->{file
}, 1)) {
3987 check_volume_storage_type
($storecfg, $drive->{file
});
3988 push @$vollist, $drive->{file
};
3991 # ignore efidisk here, already added in bios/fw handling code above
3992 return if $drive->{interface
} eq 'efidisk';
3994 return if $drive->{interface
} eq 'tpmstate';
3996 $use_virtio = 1 if $ds =~ m/^virtio/;
3998 $drive->{bootindex
} = $bootorder->{$ds} if $bootorder->{$ds};
4000 if ($drive->{interface
} eq 'virtio'){
4001 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread
};
4004 if ($drive->{interface
} eq 'scsi') {
4006 my ($maxdev, $controller, $controller_prefix) = scsihw_infos
($conf, $drive);
4008 die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
4009 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
4011 my $pciaddr = print_pci_addr
("$controller_prefix$controller", $bridges, $arch, $machine_type);
4012 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ?
"virtio-scsi-pci" : $scsihw;
4015 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{iothread
}){
4016 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4017 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4018 } elsif ($drive->{iothread
}) {
4020 "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n"
4025 if($conf->{scsihw
} && $conf->{scsihw
} eq "virtio-scsi-single" && $drive->{queues
}){
4026 $queues = ",num_queues=$drive->{queues}";
4029 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues"
4030 if !$scsicontroller->{$controller};
4031 $scsicontroller->{$controller}=1;
4034 if ($drive->{interface
} eq 'sata') {
4035 my $controller = int($drive->{index} / $PVE::QemuServer
::Drive
::MAX_SATA_DISKS
);
4036 my $pciaddr = print_pci_addr
("ahci$controller", $bridges, $arch, $machine_type);
4037 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr"
4038 if !$ahcicontroller->{$controller};
4039 $ahcicontroller->{$controller}=1;
4042 my $pbs_conf = $pbs_backing->{$ds};
4043 my $pbs_name = undef;
4045 $pbs_name = "drive-$ds-pbs";
4046 push @$devices, '-blockdev', print_pbs_blockdev
($pbs_conf, $pbs_name);
4049 my $drive_cmd = print_drive_commandline_full
(
4050 $storecfg, $vmid, $drive, $pbs_name, min_version
($kvmver, 6, 0));
4052 # extra protection for templates, but SATA and IDE don't support it..
4053 $drive_cmd .= ',readonly=on' if drive_is_read_only
($conf, $drive);
4055 push @$devices, '-drive',$drive_cmd;
4056 push @$devices, '-device', print_drivedevice_full
(
4057 $storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4060 for (my $i = 0; $i < $MAX_NETS; $i++) {
4061 my $netname = "net$i";
4063 next if !$conf->{$netname};
4064 my $d = parse_net
($conf->{$netname});
4066 # save the MAC addr here (could be auto-gen. in some odd setups) for FDB registering later?
4068 $use_virtio = 1 if $d->{model
} eq 'virtio';
4070 $d->{bootindex
} = $bootorder->{$netname} if $bootorder->{$netname};
4072 my $netdevfull = print_netdev_full
($vmid, $conf, $arch, $d, $netname);
4073 push @$devices, '-netdev', $netdevfull;
4075 my $netdevicefull = print_netdevice_full
(
4076 $vmid, $conf, $d, $netname, $bridges, $use_old_bios_files, $arch, $machine_type, $machine_version);
4078 push @$devices, '-device', $netdevicefull;
4081 if ($conf->{ivshmem
}) {
4082 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
4086 $bus = print_pcie_addr
("ivshmem");
4088 $bus = print_pci_addr
("ivshmem", $bridges, $arch, $machine_type);
4091 my $ivshmem_name = $ivshmem->{name
} // $vmid;
4092 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4094 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4095 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path"
4096 .",size=$ivshmem->{size}M";
4099 # pci.4 is nested in pci.1
4100 $bridges->{1} = 1 if $bridges->{4};
4102 if (!$q35) { # add pci bridges
4103 if (min_version
($machine_version, 2, 3)) {
4107 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4110 for my $k (sort {$b cmp $a} keys %$bridges) {
4111 next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
4114 if ($k == 2 && $legacy_igd) {
4117 my $pciaddr = print_pci_addr
("pci.$k_name", undef, $arch, $machine_type);
4118 my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
4120 if ($q35) { # add after -readconfig pve-q35.cfg
4121 splice @$devices, 2, 0, '-device', $devstr;
4123 unshift @$devices, '-device', $devstr if $k > 0;
4128 push @$machineFlags, 'accel=tcg';
4131 push @$machineFlags, 'smm=off' if should_disable_smm
($conf, $vga);
4133 my $machine_type_min = $machine_type;
4134 if ($add_pve_version) {
4135 $machine_type_min =~ s/\+pve\d+$//;
4136 $machine_type_min .= "+pve$required_pve_version";
4138 push @$machineFlags, "type=${machine_type_min}";
4140 push @$cmd, @$devices;
4141 push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
4142 push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
4143 push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
4145 if (my $vmstate = $conf->{vmstate
}) {
4146 my $statepath = PVE
::Storage
::path
($storecfg, $vmstate);
4147 push @$vollist, $vmstate;
4148 push @$cmd, '-loadstate', $statepath;
4149 print "activating and using '$vmstate' as vmstate\n";
4152 if (PVE
::QemuConfig-
>is_template($conf)) {
4153 # needed to workaround base volumes being read-only
4154 push @$cmd, '-snapshot';
4158 if ($conf->{args
}) {
4159 my $aa = PVE
::Tools
::split_args
($conf->{args
});
4163 return wantarray ?
($cmd, $vollist, $spice_port) : $cmd;
4166 sub check_rng_source
{
4169 # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
4170 die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
4173 my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
4174 if ($source eq '/dev/hwrng' && file_read_firstline
($rng_current) eq 'none') {
4175 # Needs to abort, otherwise QEMU crashes on first rng access. Note that rng_current cannot
4176 # be changed to 'none' manually, so once the VM is past this point, it's no longer an issue.
4177 die "Cannot start VM with passed-through RNG device: '/dev/hwrng' exists, but"
4178 ." '$rng_current' is set to 'none'. Ensure that a compatible hardware-RNG is attached"
4186 my $res = mon_cmd
($vmid, 'query-spice');
4188 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4191 sub vm_devices_list
{
4194 my $res = mon_cmd
($vmid, 'query-pci');
4195 my $devices_to_check = [];
4197 foreach my $pcibus (@$res) {
4198 push @$devices_to_check, @{$pcibus->{devices
}},
4201 while (@$devices_to_check) {
4203 for my $d (@$devices_to_check) {
4204 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4205 next if !$d->{'pci_bridge'} || !$d->{'pci_bridge'}->{devices
};
4207 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices
}});
4208 push @$to_check, @{$d->{'pci_bridge'}->{devices
}};
4210 $devices_to_check = $to_check;
4213 my $resblock = mon_cmd
($vmid, 'query-block');
4214 foreach my $block (@$resblock) {
4215 if($block->{device
} =~ m/^drive-(\S+)/){
4220 my $resmice = mon_cmd
($vmid, 'query-mice');
4221 foreach my $mice (@$resmice) {
4222 if ($mice->{name
} eq 'QEMU HID Tablet') {
4223 $devices->{tablet
} = 1;
4228 # for usb devices there is no query-usb
4229 # but we can iterate over the entries in
4230 # qom-list path=/machine/peripheral
4231 my $resperipheral = mon_cmd
($vmid, 'qom-list', path
=> '/machine/peripheral');
4232 foreach my $per (@$resperipheral) {
4233 if ($per->{name
} =~ m/^usb(?:redirdev)?\d+$/) {
4234 $devices->{$per->{name
}} = 1;
4242 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4244 my $q35 = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
4246 my $devices_list = vm_devices_list
($vmid);
4247 return 1 if defined($devices_list->{$deviceid});
4249 # add PCI bridge if we need it for the device
4250 qemu_add_pci_bridge
($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type);
4252 if ($deviceid eq 'tablet') {
4253 qemu_deviceadd
($vmid, print_tabletdevice_full
($conf, $arch));
4254 } elsif ($deviceid eq 'keyboard') {
4255 qemu_deviceadd
($vmid, print_keyboarddevice_full
($conf, $arch));
4256 } elsif ($deviceid =~ m/^usbredirdev(\d+)$/) {
4258 qemu_spice_usbredir_chardev_add
($vmid, "usbredirchardev$id");
4259 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_spice_usbdevice
($id, "xhci", $id + 1));
4260 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4261 qemu_deviceadd
($vmid, PVE
::QemuServer
::USB
::print_usbdevice_full
($conf, $deviceid, $device, {}, $1 + 1));
4262 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4263 qemu_iothread_add
($vmid, $deviceid, $device);
4265 qemu_driveadd
($storecfg, $vmid, $device);
4266 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4268 qemu_deviceadd
($vmid, $devicefull);
4269 eval { qemu_deviceaddverify
($vmid, $deviceid); };
4271 eval { qemu_drivedel
($vmid, $deviceid); };
4275 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4276 my $scsihw = defined($conf->{scsihw
}) ?
$conf->{scsihw
} : "lsi";
4277 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4278 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ?
"virtio-scsi-pci" : $scsihw;
4280 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4282 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread
}) {
4283 qemu_iothread_add
($vmid, $deviceid, $device);
4284 $devicefull .= ",iothread=iothread-$deviceid";
4287 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues
}) {
4288 $devicefull .= ",num_queues=$device->{queues}";
4291 qemu_deviceadd
($vmid, $devicefull);
4292 qemu_deviceaddverify
($vmid, $deviceid);
4293 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4294 qemu_findorcreatescsihw
($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4295 qemu_driveadd
($storecfg, $vmid, $device);
4297 my $devicefull = print_drivedevice_full
($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
4298 eval { qemu_deviceadd
($vmid, $devicefull); };
4300 eval { qemu_drivedel
($vmid, $deviceid); };
4304 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4305 return if !qemu_netdevadd
($vmid, $conf, $arch, $device, $deviceid);
4307 my $machine_type = PVE
::QemuServer
::Machine
::qemu_machine_pxe
($vmid, $conf);
4308 my $machine_version = PVE
::QemuServer
::Machine
::extract_version
($machine_type);
4309 my $use_old_bios_files = undef;
4310 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files
($machine_type);
4312 my $netdevicefull = print_netdevice_full
(
4313 $vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type, $machine_version);
4314 qemu_deviceadd
($vmid, $netdevicefull);
4316 qemu_deviceaddverify
($vmid, $deviceid);
4317 qemu_set_link_status
($vmid, $deviceid, !$device->{link_down
});
4320 eval { qemu_netdevdel
($vmid, $deviceid); };
4324 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4326 my $pciaddr = print_pci_addr
($deviceid, undef, $arch, $machine_type);
4327 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4329 qemu_deviceadd
($vmid, $devicefull);
4330 qemu_deviceaddverify
($vmid, $deviceid);
4332 die "can't hotplug device '$deviceid'\n";
4338 # fixme: this should raise exceptions on error!
4339 sub vm_deviceunplug
{
4340 my ($vmid, $conf, $deviceid) = @_;
4342 my $devices_list = vm_devices_list
($vmid);
4343 return 1 if !defined($devices_list->{$deviceid});
4345 my $bootdisks = PVE
::QemuServer
::Drive
::get_bootdisks
($conf);
4346 die "can't unplug bootdisk '$deviceid'\n" if grep {$_ eq $deviceid} @$bootdisks;
4348 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard' || $deviceid eq 'xhci') {
4349 qemu_devicedel
($vmid, $deviceid);
4350 } elsif ($deviceid =~ m/^usbredirdev\d+$/) {
4351 qemu_devicedel
($vmid, $deviceid);
4352 qemu_devicedelverify
($vmid, $deviceid);
4353 } elsif ($deviceid =~ m/^usb\d+$/) {
4354 qemu_devicedel
($vmid, $deviceid);
4355 qemu_devicedelverify
($vmid, $deviceid);
4356 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4357 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4359 qemu_devicedel
($vmid, $deviceid);
4360 qemu_devicedelverify
($vmid, $deviceid);
4361 qemu_drivedel
($vmid, $deviceid);
4362 qemu_iothread_del
($vmid, $deviceid, $device);
4363 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4364 qemu_devicedel
($vmid, $deviceid);
4365 qemu_devicedelverify
($vmid, $deviceid);
4366 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4367 my $device = parse_drive
($deviceid, $conf->{$deviceid});
4369 qemu_devicedel
($vmid, $deviceid);
4370 qemu_devicedelverify
($vmid, $deviceid);
4371 qemu_drivedel
($vmid, $deviceid);
4372 qemu_deletescsihw
($conf, $vmid, $deviceid);
4374 qemu_iothread_del
($vmid, "virtioscsi$device->{index}", $device)
4375 if $conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single');
4376 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4377 qemu_devicedel
($vmid, $deviceid);
4378 qemu_devicedelverify
($vmid, $deviceid);
4379 qemu_netdevdel
($vmid, $deviceid);
4381 die "can't unplug device '$deviceid'\n";
4387 sub qemu_spice_usbredir_chardev_add
{
4388 my ($vmid, $id) = @_;
4390 mon_cmd
($vmid, "chardev-add" , (
4401 sub qemu_deviceadd
{
4402 my ($vmid, $devicefull) = @_;
4404 $devicefull = "driver=".$devicefull;
4405 my %options = split(/[=,]/, $devicefull);
4407 mon_cmd
($vmid, "device_add" , %options);
4410 sub qemu_devicedel
{
4411 my ($vmid, $deviceid) = @_;
4413 my $ret = mon_cmd
($vmid, "device_del", id
=> $deviceid);
4416 sub qemu_iothread_add
{
4417 my ($vmid, $deviceid, $device) = @_;
4419 if ($device->{iothread
}) {
4420 my $iothreads = vm_iothreads_list
($vmid);
4421 qemu_objectadd
($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4425 sub qemu_iothread_del
{
4426 my ($vmid, $deviceid, $device) = @_;
4428 if ($device->{iothread
}) {
4429 my $iothreads = vm_iothreads_list
($vmid);
4430 qemu_objectdel
($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4434 sub qemu_objectadd
{
4435 my ($vmid, $objectid, $qomtype) = @_;
4437 mon_cmd
($vmid, "object-add", id
=> $objectid, "qom-type" => $qomtype);
4442 sub qemu_objectdel
{
4443 my ($vmid, $objectid) = @_;
4445 mon_cmd
($vmid, "object-del", id
=> $objectid);
4451 my ($storecfg, $vmid, $device) = @_;
4453 my $kvmver = get_running_qemu_version
($vmid);
4454 my $io_uring = min_version
($kvmver, 6, 0);
4455 my $drive = print_drive_commandline_full
($storecfg, $vmid, $device, undef, $io_uring);
4456 $drive =~ s/\\/\\\\/g;
4457 my $ret = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "drive_add auto \"$drive\"");
4459 # If the command succeeds qemu prints: "OK
"
4460 return 1 if $ret =~ m/OK/s;
4462 die "adding drive failed
: $ret\n";
4466 my ($vmid, $deviceid) = @_;
4468 my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-
$deviceid");
4471 return 1 if $ret eq "";
4473 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4474 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4476 die "deleting drive
$deviceid failed
: $ret\n";
4479 sub qemu_deviceaddverify {
4480 my ($vmid, $deviceid) = @_;
4482 for (my $i = 0; $i <= 5; $i++) {
4483 my $devices_list = vm_devices_list($vmid);
4484 return 1 if defined($devices_list->{$deviceid});
4488 die "error on hotplug device
'$deviceid'\n";
4492 sub qemu_devicedelverify {
4493 my ($vmid, $deviceid) = @_;
4495 # need to verify that the device is correctly removed as device_del
4496 # is async and empty return is not reliable
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 hot-unplugging device
'$deviceid'\n";
4507 sub qemu_findorcreatescsihw {
4508 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4510 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4512 my $scsihwid="$controller_prefix$controller";
4513 my $devices_list = vm_devices_list($vmid);
4515 if (!defined($devices_list->{$scsihwid})) {
4516 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4522 sub qemu_deletescsihw {
4523 my ($conf, $vmid, $opt) = @_;
4525 my $device = parse_drive($opt, $conf->{$opt});
4527 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4528 vm_deviceunplug($vmid, $conf, "virtioscsi
$device->{index}");
4532 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4534 my $devices_list = vm_devices_list($vmid);
4535 foreach my $opt (keys %{$devices_list}) {
4536 if (is_valid_drivename($opt)) {
4537 my $drive = parse_drive($opt, $conf->{$opt});
4538 if ($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4544 my $scsihwid="scsihw
$controller";
4546 vm_deviceunplug($vmid, $conf, $scsihwid);
4551 sub qemu_add_pci_bridge {
4552 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4558 print_pci_addr($device, $bridges, $arch, $machine_type);
4560 while (my ($k, $v) = each %$bridges) {
4563 return 1 if !defined($bridgeid) || $bridgeid < 1;
4565 my $bridge = "pci
.$bridgeid";
4566 my $devices_list = vm_devices_list($vmid);
4568 if (!defined($devices_list->{$bridge})) {
4569 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4575 sub qemu_set_link_status {
4576 my ($vmid, $device, $up) = @_;
4578 mon_cmd($vmid, "set_link
", name => $device,
4579 up => $up ? JSON::true : JSON::false);
4582 sub qemu_netdevadd {
4583 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4585 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4586 my %options = split(/[=,]/, $netdev);
4588 if (defined(my $vhost = $options{vhost})) {
4589 $options{vhost} = JSON::boolean(PVE::JSONSchema::parse_boolean($vhost));
4592 if (defined(my $queues = $options{queues})) {
4593 $options{queues} = $queues + 0;
4596 mon_cmd($vmid, "netdev_add
", %options);
4600 sub qemu_netdevdel {
4601 my ($vmid, $deviceid) = @_;
4603 mon_cmd($vmid, "netdev_del
", id => $deviceid);
4606 sub qemu_usb_hotplug {
4607 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4611 # remove the old one first
4612 vm_deviceunplug($vmid, $conf, $deviceid);
4614 # check if xhci controller is necessary and available
4615 my $devicelist = vm_devices_list($vmid);
4617 if (!$devicelist->{xhci}) {
4618 my $pciaddr = print_pci_addr("xhci
", undef, $arch, $machine_type);
4619 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_qemu_xhci_controller($pciaddr));
4622 # print_usbdevice_full expects the parsed device
4623 my $d = parse_usb_device($device->{host});
4624 $d->{usb3} = $device->{usb3};
4627 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4630 sub qemu_cpu_hotplug {
4631 my ($vmid, $conf, $vcpus) = @_;
4633 my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4636 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4637 $sockets = $conf->{sockets} if $conf->{sockets};
4638 my $cores = $conf->{cores} || 1;
4639 my $maxcpus = $sockets * $cores;
4641 $vcpus = $maxcpus if !$vcpus;
4643 die "you can
't add more vcpus than maxcpus\n"
4644 if $vcpus > $maxcpus;
4646 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4648 if ($vcpus < $currentvcpus) {
4650 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4652 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4653 qemu_devicedel($vmid, "cpu$i");
4655 my $currentrunningvcpus = undef;
4657 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4658 last if scalar(@{$currentrunningvcpus}) == $i-1;
4659 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4663 #update conf after each succesfull cpu unplug
4664 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4665 PVE::QemuConfig->write_config($vmid, $conf);
4668 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4674 my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4675 die "vcpus in running vm does not match its configuration\n"
4676 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4678 if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4680 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4681 my $cpustr = print_cpu_device($conf, $i);
4682 qemu_deviceadd($vmid, $cpustr);
4685 my $currentrunningvcpus = undef;
4687 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4688 last if scalar(@{$currentrunningvcpus}) == $i;
4689 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4693 #update conf after each succesfull cpu hotplug
4694 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4695 PVE::QemuConfig->write_config($vmid, $conf);
4699 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4700 mon_cmd($vmid, "cpu-add", id => int($i));
4705 sub qemu_block_set_io_throttle {
4706 my ($vmid, $deviceid,
4707 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4708 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4709 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4710 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4712 return if !check_running($vmid) ;
4714 mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4716 bps_rd => int($bps_rd),
4717 bps_wr => int($bps_wr),
4719 iops_rd => int($iops_rd),
4720 iops_wr => int($iops_wr),
4721 bps_max => int($bps_max),
4722 bps_rd_max => int($bps_rd_max),
4723 bps_wr_max => int($bps_wr_max),
4724 iops_max => int($iops_max),
4725 iops_rd_max => int($iops_rd_max),
4726 iops_wr_max => int($iops_wr_max),
4727 bps_max_length => int($bps_max_length),
4728 bps_rd_max_length => int($bps_rd_max_length),
4729 bps_wr_max_length => int($bps_wr_max_length),
4730 iops_max_length => int($iops_max_length),
4731 iops_rd_max_length => int($iops_rd_max_length),
4732 iops_wr_max_length => int($iops_wr_max_length),
4737 sub qemu_block_resize {
4738 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4740 my $running = check_running($vmid);
4742 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4744 return if !$running;
4746 my $padding = (1024 - $size % 1024) % 1024;
4747 $size = $size + $padding;
4752 device => $deviceid,
4758 sub qemu_volume_snapshot {
4759 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4761 my $running = check_running($vmid);
4763 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4764 mon_cmd($vmid, 'blockdev-snapshot-internal-sync
', device => $deviceid, name => $snap);
4766 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4770 sub qemu_volume_snapshot_delete {
4771 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4773 my $running = check_running($vmid);
4778 my $conf = PVE::QemuConfig->load_config($vmid);
4779 PVE::QemuConfig->foreach_volume($conf, sub {
4780 my ($ds, $drive) = @_;
4781 $running = 1 if $drive->{file} eq $volid;
4785 if ($running && do_snapshots_with_qemu($storecfg, $volid, $deviceid)) {
4786 mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync
', device => $deviceid, name => $snap);
4788 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4792 sub set_migration_caps {
4793 my ($vmid, $savevm) = @_;
4795 my $qemu_support = eval { mon_cmd($vmid, "query-proxmox-support") };
4797 my $bitmap_prop = $savevm ? 'pbs-dirty-bitmap-savevm
' : 'pbs-dirty-bitmap-migration
';
4798 my $dirty_bitmaps = $qemu_support->{$bitmap_prop} ? 1 : 0;
4803 "auto-converge" => 1,
4805 "x-rdma-pin-all" => 0,
4808 "dirty-bitmaps" => $dirty_bitmaps,
4811 my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4813 for my $supported_capability (@$supported_capabilities) {
4815 capability => $supported_capability->{capability},
4816 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4820 mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4824 my ($conf, $func, @param) = @_;
4828 my $test_volid = sub {
4829 my ($key, $drive, $snapname) = @_;
4831 my $volid = $drive->{file};
4834 $volhash->{$volid}->{cdrom} //= 1;
4835 $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4837 my $replicate = $drive->{replicate} // 1;
4838 $volhash->{$volid}->{replicate} //= 0;
4839 $volhash->{$volid}->{replicate} = 1 if $replicate;
4841 $volhash->{$volid}->{shared} //= 0;
4842 $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4844 $volhash->{$volid}->{referenced_in_config} //= 0;
4845 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
4847 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4848 if defined($snapname);
4850 my $size = $drive->{size};
4851 $volhash->{$volid}->{size} //= $size if $size;
4853 $volhash->{$volid}->{is_vmstate} //= 0;
4854 $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate
';
4856 $volhash->{$volid}->{is_tpmstate} //= 0;
4857 $volhash->{$volid}->{is_tpmstate} = 1 if $key eq 'tpmstate0
';
4859 $volhash->{$volid}->{is_unused} //= 0;
4860 $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4862 $volhash->{$volid}->{drivename} = $key if is_valid_drivename($key);
4865 my $include_opts = {
4866 extra_keys => ['vmstate
'],
4867 include_unused => 1,
4870 PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4871 foreach my $snapname (keys %{$conf->{snapshots}}) {
4872 my $snap = $conf->{snapshots}->{$snapname};
4873 PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4876 foreach my $volid (keys %$volhash) {
4877 &$func($volid, $volhash->{$volid}, @param);
4881 my $fast_plug_option = {
4889 'vmstatestorage
' => 1,
4894 for my $opt (keys %$confdesc_cloudinit) {
4895 $fast_plug_option->{$opt} = 1;
4898 # hotplug changes in [PENDING]
4899 # $selection hash can be used to only apply specified options, for
4900 # example: { cores => 1 } (only apply changed 'cores
')
4901 # $errors ref is used to return error messages
4902 sub vmconfig_hotplug_pending {
4903 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4905 my $defaults = load_defaults();
4906 my $arch = get_vm_arch($conf);
4907 my $machine_type = get_vm_machine($conf, undef, $arch);
4909 # commit values which do not have any impact on running VM first
4910 # Note: those option cannot raise errors, we we do not care about
4911 # $selection and always apply them.
4913 my $add_error = sub {
4914 my ($opt, $msg) = @_;
4915 $errors->{$opt} = "hotplug problem - $msg";
4918 my $cloudinit_pending_properties = PVE::QemuServer::cloudinit_pending_properties();
4920 my $cloudinit_record_changed = sub {
4921 my ($conf, $opt, $old, $new) = @_;
4922 return if !$cloudinit_pending_properties->{$opt};
4924 my $ci = ($conf->{cloudinit} //= {});
4926 my $recorded = $ci->{$opt};
4927 my %added = map { $_ => 1 } PVE::Tools::split_list(delete($ci->{added}) // '');
4929 if (defined($new)) {
4930 if (defined($old)) {
4931 # an existing value is being modified
4932 if (defined($recorded)) {
4933 # the value was already not in sync
4934 if ($new eq $recorded) {
4935 # a value is being reverted to the cloud-init state:
4937 delete $added{$opt};
4939 # the value was changed multiple times, do nothing
4941 } elsif ($added{$opt}) {
4942 # the value had been marked as added and is being changed, do nothing
4944 # the value is new, record it:
4948 # a new value is being added
4949 if (defined($recorded)) {
4950 # it was already not in sync
4951 if ($new eq $recorded) {
4952 # a value is being reverted to the cloud-init state:
4954 delete $added{$opt};
4956 # the value had temporarily been removed, do nothing
4958 } elsif ($added{$opt}) {
4959 # the value had been marked as added already, do nothing
4961 # the value is new, add it
4965 } elsif (!defined($old)) {
4966 # a non-existent value is being removed? ignore...
4968 # a value is being deleted
4969 if (defined($recorded)) {
4970 # a value was already recorded, just keep it
4971 } elsif ($added{$opt}) {
4972 # the value was marked as added, remove it
4973 delete $added{$opt};
4975 # a previously unrecorded value is being removed, record the old value:
4980 my $added = join(',', sort keys %added);
4981 $ci->{added} = $added if length($added);
4985 foreach my $opt (keys %{$conf->{pending}}) { # add/change
4986 if ($fast_plug_option->{$opt}) {
4987 my $new = delete $conf->{pending}->{$opt};
4988 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $new);
4989 $conf->{$opt} = $new;
4995 PVE::QemuConfig->write_config($vmid, $conf);
4998 my $ostype = $conf->{ostype};
4999 my $version = extract_version($machine_type, get_running_qemu_version($vmid));
5000 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
5001 my $usb_hotplug = $hotplug_features->{usb}
5002 && min_version($version, 7, 1)
5003 && defined($ostype) && ($ostype eq 'l26
' || windows_version($ostype) > 7);
5005 my $cgroup = PVE::QemuServer::CGroup->new($vmid);
5006 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
5008 foreach my $opt (sort keys %$pending_delete_hash) {
5009 next if $selection && !$selection->{$opt};
5010 my $force = $pending_delete_hash->{$opt}->{force};
5012 if ($opt eq 'hotplug
') {
5013 die "skip\n" if ($conf->{hotplug} =~ /memory/);
5014 } elsif ($opt eq 'tablet
') {
5015 die "skip\n" if !$hotplug_features->{usb};
5016 if ($defaults->{tablet}) {
5017 vm_deviceplug($storecfg, $conf, $vmid, 'tablet
', $arch, $machine_type);
5018 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard
', $arch, $machine_type)
5019 if $arch eq 'aarch64
';
5021 vm_deviceunplug($vmid, $conf, 'tablet
');
5022 vm_deviceunplug($vmid, $conf, 'keyboard
') if $arch eq 'aarch64
';
5024 } elsif ($opt =~ m/^usb(\d+)$/) {
5026 die "skip\n" if !$usb_hotplug;
5027 vm_deviceunplug($vmid, $conf, "usbredirdev$index"); # if it's a spice port
5028 vm_deviceunplug
($vmid, $conf, $opt);
5029 } elsif ($opt eq 'vcpus') {
5030 die "skip\n" if !$hotplug_features->{cpu
};
5031 qemu_cpu_hotplug
($vmid, $conf, undef);
5032 } elsif ($opt eq 'balloon') {
5033 # enable balloon device is not hotpluggable
5034 die "skip\n" if defined($conf->{balloon
}) && $conf->{balloon
} == 0;
5035 # here we reset the ballooning value to memory
5036 my $balloon = $conf->{memory
} || $defaults->{memory
};
5037 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5038 } elsif ($fast_plug_option->{$opt}) {
5040 } elsif ($opt =~ m/^net(\d+)$/) {
5041 die "skip\n" if !$hotplug_features->{network
};
5042 vm_deviceunplug
($vmid, $conf, $opt);
5043 } elsif (is_valid_drivename
($opt)) {
5044 die "skip\n" if !$hotplug_features->{disk
} || $opt =~ m/(ide|sata)(\d+)/;
5045 vm_deviceunplug
($vmid, $conf, $opt);
5046 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5047 } elsif ($opt =~ m/^memory$/) {
5048 die "skip\n" if !$hotplug_features->{memory
};
5049 PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults, $opt);
5050 } elsif ($opt eq 'cpuunits') {
5051 $cgroup->change_cpu_shares(undef);
5052 } elsif ($opt eq 'cpulimit') {
5053 $cgroup->change_cpu_quota(undef, undef); # reset, cgroup module can better decide values
5059 &$add_error($opt, $err) if $err ne "skip\n";
5061 my $old = delete $conf->{$opt};
5062 $cloudinit_record_changed->($conf, $opt, $old, undef);
5063 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5068 foreach my $opt (keys %{$conf->{pending
}}) {
5069 next if $selection && !$selection->{$opt};
5070 my $value = $conf->{pending
}->{$opt};
5072 if ($opt eq 'hotplug') {
5073 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug
} =~ /memory/);
5074 } elsif ($opt eq 'tablet') {
5075 die "skip\n" if !$hotplug_features->{usb
};
5077 vm_deviceplug
($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5078 vm_deviceplug
($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5079 if $arch eq 'aarch64';
5080 } elsif ($value == 0) {
5081 vm_deviceunplug
($vmid, $conf, 'tablet');
5082 vm_deviceunplug
($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5084 } elsif ($opt =~ m/^usb(\d+)$/) {
5086 die "skip\n" if !$usb_hotplug;
5087 my $d = eval { parse_property_string
($usbdesc->{format
}, $value) };
5089 if ($d->{host
} eq 'spice') {
5090 $id = "usbredirdev$index";
5092 qemu_usb_hotplug
($storecfg, $conf, $vmid, $id, $d, $arch, $machine_type);
5093 } elsif ($opt eq 'vcpus') {
5094 die "skip\n" if !$hotplug_features->{cpu
};
5095 qemu_cpu_hotplug
($vmid, $conf, $value);
5096 } elsif ($opt eq 'balloon') {
5097 # enable/disable balloning device is not hotpluggable
5098 my $old_balloon_enabled = !!(!defined($conf->{balloon
}) || $conf->{balloon
});
5099 my $new_balloon_enabled = !!(!defined($conf->{pending
}->{balloon
}) || $conf->{pending
}->{balloon
});
5100 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5102 # allow manual ballooning if shares is set to zero
5103 if ((defined($conf->{shares
}) && ($conf->{shares
} == 0))) {
5104 my $balloon = $conf->{pending
}->{balloon
} || $conf->{memory
} || $defaults->{memory
};
5105 mon_cmd
($vmid, "balloon", value
=> $balloon*1024*1024);
5107 } elsif ($opt =~ m/^net(\d+)$/) {
5108 # some changes can be done without hotplug
5109 vmconfig_update_net
($storecfg, $conf, $hotplug_features->{network
},
5110 $vmid, $opt, $value, $arch, $machine_type);
5111 } elsif (is_valid_drivename
($opt)) {
5112 die "skip\n" if $opt eq 'efidisk0' || $opt eq 'tpmstate0';
5113 # some changes can be done without hotplug
5114 my $drive = parse_drive
($opt, $value);
5115 if (drive_is_cloudinit
($drive)) {
5116 $cloudinit_opt = [$opt, $drive];
5117 # apply all the other changes first, then generate the cloudinit disk
5120 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5121 $vmid, $opt, $value, $arch, $machine_type);
5122 } elsif ($opt =~ m/^memory$/) { #dimms
5123 die "skip\n" if !$hotplug_features->{memory
};
5124 $value = PVE
::QemuServer
::Memory
::qemu_memory_hotplug
($vmid, $conf, $defaults, $opt, $value);
5125 } elsif ($opt eq 'cpuunits') {
5126 my $new_cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{pending
}->{$opt}); #clamp
5127 $cgroup->change_cpu_shares($new_cpuunits);
5128 } elsif ($opt eq 'cpulimit') {
5129 my $cpulimit = $conf->{pending
}->{$opt} == 0 ?
-1 : int($conf->{pending
}->{$opt} * 100000);
5130 $cgroup->change_cpu_quota($cpulimit, 100000);
5131 } elsif ($opt eq 'agent') {
5132 vmconfig_update_agent
($conf, $opt, $value);
5134 die "skip\n"; # skip non-hot-pluggable options
5138 &$add_error($opt, $err) if $err ne "skip\n";
5140 $cloudinit_record_changed->($conf, $opt, $conf->{$opt}, $value);
5141 $conf->{$opt} = $value;
5142 delete $conf->{pending
}->{$opt};
5146 if (defined($cloudinit_opt)) {
5147 my ($opt, $drive) = @$cloudinit_opt;
5148 my $value = $conf->{pending
}->{$opt};
5150 my $temp = {%$conf, $opt => $value};
5151 PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($temp, $vmid);
5152 vmconfig_update_disk
($storecfg, $conf, $hotplug_features->{disk
},
5153 $vmid, $opt, $value, $arch, $machine_type);
5156 &$add_error($opt, $err) if $err ne "skip\n";
5158 $conf->{$opt} = $value;
5159 delete $conf->{pending
}->{$opt};
5163 # unplug xhci controller if no usb device is left
5166 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
5167 next if !defined($conf->{"usb$i"});
5172 vm_deviceunplug
($vmid, $conf, 'xhci');
5176 PVE
::QemuConfig-
>write_config($vmid, $conf);
5178 if ($hotplug_features->{cloudinit
} && PVE
::QemuServer
::Cloudinit
::has_changes
($conf)) {
5179 PVE
::QemuServer
::vmconfig_update_cloudinit_drive
($storecfg, $conf, $vmid);
5183 sub try_deallocate_drive
{
5184 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5186 if (($force || $key =~ /^unused/) && !drive_is_cdrom
($drive, 1)) {
5187 my $volid = $drive->{file
};
5188 if (vm_is_volid_owner
($storecfg, $vmid, $volid)) {
5189 my $sid = PVE
::Storage
::parse_volume_id
($volid);
5190 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5192 # check if the disk is really unused
5193 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5194 if PVE
::QemuServer
::Drive
::is_volume_in_use
($storecfg, $conf, $key, $volid);
5195 PVE
::Storage
::vdisk_free
($storecfg, $volid);
5198 # If vm is not owner of this disk remove from config
5206 sub vmconfig_delete_or_detach_drive
{
5207 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5209 my $drive = parse_drive
($opt, $conf->{$opt});
5211 my $rpcenv = PVE
::RPCEnvironment
::get
();
5212 my $authuser = $rpcenv->get_user();
5215 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5216 try_deallocate_drive
($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5218 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $drive);
5224 sub vmconfig_apply_pending
{
5225 my ($vmid, $conf, $storecfg, $errors, $skip_cloud_init) = @_;
5227 return if !scalar(keys %{$conf->{pending
}});
5229 my $add_apply_error = sub {
5230 my ($opt, $msg) = @_;
5231 my $err_msg = "unable to apply pending change $opt : $msg";
5232 $errors->{$opt} = $err_msg;
5238 my $pending_delete_hash = PVE
::QemuConfig-
>parse_pending_delete($conf->{pending
}->{delete});
5239 foreach my $opt (sort keys %$pending_delete_hash) {
5240 my $force = $pending_delete_hash->{$opt}->{force
};
5242 if ($opt =~ m/^unused/) {
5243 die "internal error";
5244 } elsif (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5245 vmconfig_delete_or_detach_drive
($vmid, $storecfg, $conf, $opt, $force);
5249 $add_apply_error->($opt, $err);
5251 PVE
::QemuConfig-
>remove_from_pending_delete($conf, $opt);
5252 delete $conf->{$opt};
5256 PVE
::QemuConfig-
>cleanup_pending($conf);
5258 my $generate_cloudinit = $skip_cloud_init ?
0 : undef;
5260 foreach my $opt (keys %{$conf->{pending
}}) { # add/change
5261 next if $opt eq 'delete'; # just to be sure
5263 if (defined($conf->{$opt}) && is_valid_drivename
($opt)) {
5264 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, parse_drive
($opt, $conf->{$opt}))
5268 $add_apply_error->($opt, $err);
5271 if (is_valid_drivename
($opt)) {
5272 my $drive = parse_drive
($opt, $conf->{pending
}->{$opt});
5273 $generate_cloudinit //= 1 if drive_is_cloudinit
($drive);
5276 $conf->{$opt} = delete $conf->{pending
}->{$opt};
5280 # write all changes at once to avoid unnecessary i/o
5281 PVE
::QemuConfig-
>write_config($vmid, $conf);
5282 if ($generate_cloudinit) {
5283 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5284 # After successful generation and if there were changes to be applied, update the
5285 # config to drop the {cloudinit} entry.
5286 PVE
::QemuConfig-
>write_config($vmid, $conf);
5291 sub vmconfig_update_net
{
5292 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5294 my $newnet = parse_net
($value);
5296 if ($conf->{$opt}) {
5297 my $oldnet = parse_net
($conf->{$opt});
5299 if (safe_string_ne
($oldnet->{model
}, $newnet->{model
}) ||
5300 safe_string_ne
($oldnet->{macaddr
}, $newnet->{macaddr
}) ||
5301 safe_num_ne
($oldnet->{queues
}, $newnet->{queues
}) ||
5302 !($newnet->{bridge
} && $oldnet->{bridge
})) { # bridge/nat mode change
5304 # for non online change, we try to hot-unplug
5305 die "skip\n" if !$hotplug;
5306 vm_deviceunplug
($vmid, $conf, $opt);
5309 die "internal error" if $opt !~ m/net(\d+)/;
5310 my $iface = "tap${vmid}i$1";
5312 if (safe_string_ne
($oldnet->{bridge
}, $newnet->{bridge
}) ||
5313 safe_num_ne
($oldnet->{tag
}, $newnet->{tag
}) ||
5314 safe_string_ne
($oldnet->{trunks
}, $newnet->{trunks
}) ||
5315 safe_num_ne
($oldnet->{firewall
}, $newnet->{firewall
})) {
5316 PVE
::Network
::tap_unplug
($iface);
5319 PVE
::Network
::SDN
::Zones
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5321 PVE
::Network
::tap_plug
($iface, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
5323 } elsif (safe_num_ne
($oldnet->{rate
}, $newnet->{rate
})) {
5324 # Rate can be applied on its own but any change above needs to
5325 # include the rate in tap_plug since OVS resets everything.
5326 PVE
::Network
::tap_rate_limit
($iface, $newnet->{rate
});
5329 if (safe_string_ne
($oldnet->{link_down
}, $newnet->{link_down
})) {
5330 qemu_set_link_status
($vmid, $opt, !$newnet->{link_down
});
5338 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5344 sub vmconfig_update_agent
{
5345 my ($conf, $opt, $value) = @_;
5347 die "skip\n" if !$conf->{$opt};
5349 my $hotplug_options = { fstrim_cloned_disks
=> 1 };
5351 my $old_agent = parse_guest_agent
($conf);
5352 my $agent = parse_guest_agent
({$opt => $value});
5354 for my $option (keys %$agent) { # added/changed options
5355 next if defined($hotplug_options->{$option});
5356 die "skip\n" if safe_string_ne
($agent->{$option}, $old_agent->{$option});
5359 for my $option (keys %$old_agent) { # removed options
5360 next if defined($hotplug_options->{$option});
5361 die "skip\n" if safe_string_ne
($old_agent->{$option}, $agent->{$option});
5364 return; # either no actual change (e.g., format string reordered) or just hotpluggable changes
5367 sub vmconfig_update_disk
{
5368 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5370 my $drive = parse_drive
($opt, $value);
5372 if ($conf->{$opt} && (my $old_drive = parse_drive
($opt, $conf->{$opt}))) {
5373 my $media = $drive->{media
} || 'disk';
5374 my $oldmedia = $old_drive->{media
} || 'disk';
5375 die "unable to change media type\n" if $media ne $oldmedia;
5377 if (!drive_is_cdrom
($old_drive)) {
5379 if ($drive->{file
} ne $old_drive->{file
}) {
5381 die "skip\n" if !$hotplug;
5383 # unplug and register as unused
5384 vm_deviceunplug
($vmid, $conf, $opt);
5385 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive)
5388 # update existing disk
5390 # skip non hotpluggable value
5391 if (safe_string_ne
($drive->{discard
}, $old_drive->{discard
}) ||
5392 safe_string_ne
($drive->{iothread
}, $old_drive->{iothread
}) ||
5393 safe_string_ne
($drive->{queues
}, $old_drive->{queues
}) ||
5394 safe_string_ne
($drive->{cache
}, $old_drive->{cache
}) ||
5395 safe_string_ne
($drive->{ssd
}, $old_drive->{ssd
}) ||
5396 safe_string_ne
($drive->{ro
}, $old_drive->{ro
})) {
5401 if (safe_num_ne
($drive->{mbps
}, $old_drive->{mbps
}) ||
5402 safe_num_ne
($drive->{mbps_rd
}, $old_drive->{mbps_rd
}) ||
5403 safe_num_ne
($drive->{mbps_wr
}, $old_drive->{mbps_wr
}) ||
5404 safe_num_ne
($drive->{iops
}, $old_drive->{iops
}) ||
5405 safe_num_ne
($drive->{iops_rd
}, $old_drive->{iops_rd
}) ||
5406 safe_num_ne
($drive->{iops_wr
}, $old_drive->{iops_wr
}) ||
5407 safe_num_ne
($drive->{mbps_max
}, $old_drive->{mbps_max
}) ||
5408 safe_num_ne
($drive->{mbps_rd_max
}, $old_drive->{mbps_rd_max
}) ||
5409 safe_num_ne
($drive->{mbps_wr_max
}, $old_drive->{mbps_wr_max
}) ||
5410 safe_num_ne
($drive->{iops_max
}, $old_drive->{iops_max
}) ||
5411 safe_num_ne
($drive->{iops_rd_max
}, $old_drive->{iops_rd_max
}) ||
5412 safe_num_ne
($drive->{iops_wr_max
}, $old_drive->{iops_wr_max
}) ||
5413 safe_num_ne
($drive->{bps_max_length
}, $old_drive->{bps_max_length
}) ||
5414 safe_num_ne
($drive->{bps_rd_max_length
}, $old_drive->{bps_rd_max_length
}) ||
5415 safe_num_ne
($drive->{bps_wr_max_length
}, $old_drive->{bps_wr_max_length
}) ||
5416 safe_num_ne
($drive->{iops_max_length
}, $old_drive->{iops_max_length
}) ||
5417 safe_num_ne
($drive->{iops_rd_max_length
}, $old_drive->{iops_rd_max_length
}) ||
5418 safe_num_ne
($drive->{iops_wr_max_length
}, $old_drive->{iops_wr_max_length
})) {
5420 qemu_block_set_io_throttle
(
5422 ($drive->{mbps
} || 0)*1024*1024,
5423 ($drive->{mbps_rd
} || 0)*1024*1024,
5424 ($drive->{mbps_wr
} || 0)*1024*1024,
5425 $drive->{iops
} || 0,
5426 $drive->{iops_rd
} || 0,
5427 $drive->{iops_wr
} || 0,
5428 ($drive->{mbps_max
} || 0)*1024*1024,
5429 ($drive->{mbps_rd_max
} || 0)*1024*1024,
5430 ($drive->{mbps_wr_max
} || 0)*1024*1024,
5431 $drive->{iops_max
} || 0,
5432 $drive->{iops_rd_max
} || 0,
5433 $drive->{iops_wr_max
} || 0,
5434 $drive->{bps_max_length
} || 1,
5435 $drive->{bps_rd_max_length
} || 1,
5436 $drive->{bps_wr_max_length
} || 1,
5437 $drive->{iops_max_length
} || 1,
5438 $drive->{iops_rd_max_length
} || 1,
5439 $drive->{iops_wr_max_length
} || 1,
5449 if ($drive->{file
} eq 'none') {
5450 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5451 if (drive_is_cloudinit
($old_drive)) {
5452 vmconfig_register_unused_drive
($storecfg, $vmid, $conf, $old_drive);
5455 my $path = get_iso_path
($storecfg, $vmid, $drive->{file
});
5457 # force eject if locked
5458 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$opt");
5461 mon_cmd
($vmid, "blockdev-change-medium",
5462 id
=> "$opt", filename
=> "$path");
5470 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5472 PVE
::Storage
::activate_volumes
($storecfg, [$drive->{file
}]) if $drive->{file
} !~ m
|^/dev/.+|;
5473 vm_deviceplug
($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5476 sub vmconfig_update_cloudinit_drive
{
5477 my ($storecfg, $conf, $vmid) = @_;
5479 my $cloudinit_ds = undef;
5480 my $cloudinit_drive = undef;
5482 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5483 my ($ds, $drive) = @_;
5484 if (PVE
::QemuServer
::drive_is_cloudinit
($drive)) {
5485 $cloudinit_ds = $ds;
5486 $cloudinit_drive = $drive;
5490 return if !$cloudinit_drive;
5492 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5493 PVE
::QemuConfig-
>write_config($vmid, $conf);
5496 my $running = PVE
::QemuServer
::check_running
($vmid);
5499 my $path = PVE
::Storage
::path
($storecfg, $cloudinit_drive->{file
});
5501 mon_cmd
($vmid, "eject", force
=> JSON
::true
, id
=> "$cloudinit_ds");
5502 mon_cmd
($vmid, "blockdev-change-medium", id
=> "$cloudinit_ds", filename
=> "$path");
5507 # called in locked context by incoming migration
5508 sub vm_migrate_get_nbd_disks
{
5509 my ($storecfg, $conf, $replicated_volumes) = @_;
5511 my $local_volumes = {};
5512 PVE
::QemuConfig-
>foreach_volume($conf, sub {
5513 my ($ds, $drive) = @_;
5515 return if drive_is_cdrom
($drive);
5516 return if $ds eq 'tpmstate0';
5518 my $volid = $drive->{file
};
5522 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($volid);
5524 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5525 return if $scfg->{shared
};
5527 # replicated disks re-use existing state via bitmap
5528 my $use_existing = $replicated_volumes->{$volid} ?
1 : 0;
5529 $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
5531 return $local_volumes;
5534 # called in locked context by incoming migration
5535 sub vm_migrate_alloc_nbd_disks
{
5536 my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
5539 foreach my $opt (sort keys %$source_volumes) {
5540 my ($volid, $storeid, $volname, $drive, $use_existing, $format) = @{$source_volumes->{$opt}};
5542 if ($use_existing) {
5543 $nbd->{$opt}->{drivestr
} = print_drive
($drive);
5544 $nbd->{$opt}->{volid
} = $volid;
5545 $nbd->{$opt}->{replicated
} = 1;
5549 # storage mapping + volname = regular migration
5550 # storage mapping + format = remote migration
5551 # order of precedence, filtered by whether storage supports it:
5552 # 1. explicit requested format
5553 # 2. format of current volume
5554 # 3. default format of storage
5555 if (!$storagemap->{identity
}) {
5556 $storeid = PVE
::JSONSchema
::map_id
($storagemap, $storeid);
5557 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
5558 if (!$format || !grep { $format eq $_ } @$validFormats) {
5560 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5561 my $fileFormat = qemu_img_format
($scfg, $volname);
5562 $format = $fileFormat
5563 if grep { $fileFormat eq $_ } @$validFormats;
5565 $format //= $defFormat;
5568 # can't happen for remote migration, so $volname is always defined
5569 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
5570 $format = qemu_img_format
($scfg, $volname);
5573 my $size = $drive->{size
} / 1024;
5574 my $newvolid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $format, undef, $size);
5575 my $newdrive = $drive;
5576 $newdrive->{format
} = $format;
5577 $newdrive->{file
} = $newvolid;
5578 my $drivestr = print_drive
($newdrive);
5579 $nbd->{$opt}->{drivestr
} = $drivestr;
5580 $nbd->{$opt}->{volid
} = $newvolid;
5586 # see vm_start_nolock for parameters, additionally:
5588 # storagemap = parsed storage map for allocating NBD disks
5590 my ($storecfg, $vmid, $params, $migrate_opts) = @_;
5592 return PVE
::QemuConfig-
>lock_config($vmid, sub {
5593 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migrate_opts->{migratedfrom
});
5595 die "you can't start a vm if it's a template\n"
5596 if !$params->{skiptemplate
} && PVE
::QemuConfig-
>is_template($conf);
5598 my $has_suspended_lock = PVE
::QemuConfig-
>has_lock($conf, 'suspended');
5599 my $has_backup_lock = PVE
::QemuConfig-
>has_lock($conf, 'backup');
5601 my $running = check_running
($vmid, undef, $migrate_opts->{migratedfrom
});
5603 if ($has_backup_lock && $running) {
5604 # a backup is currently running, attempt to start the guest in the
5605 # existing QEMU instance
5606 return vm_resume
($vmid);
5609 PVE
::QemuConfig-
>check_lock($conf)
5610 if !($params->{skiplock
} || $has_suspended_lock);
5612 $params->{resume
} = $has_suspended_lock || defined($conf->{vmstate
});
5614 die "VM $vmid already running\n" if $running;
5616 if (my $storagemap = $migrate_opts->{storagemap
}) {
5617 my $replicated = $migrate_opts->{replicated_volumes
};
5618 my $disks = vm_migrate_get_nbd_disks
($storecfg, $conf, $replicated);
5619 $migrate_opts->{nbd
} = vm_migrate_alloc_nbd_disks
($storecfg, $vmid, $disks, $storagemap);
5621 foreach my $opt (keys %{$migrate_opts->{nbd
}}) {
5622 $conf->{$opt} = $migrate_opts->{nbd
}->{$opt}->{drivestr
};
5626 return vm_start_nolock
($storecfg, $vmid, $conf, $params, $migrate_opts);
5632 # statefile => 'tcp', 'unix' for migration or path/volid for RAM state
5633 # skiplock => 0/1, skip checking for config lock
5634 # skiptemplate => 0/1, skip checking whether VM is template
5635 # forcemachine => to force QEMU machine (rollback/migration)
5636 # forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
5637 # timeout => in seconds
5638 # paused => start VM in paused state (backup)
5639 # resume => resume from hibernation
5650 # nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
5651 # migratedfrom => source node
5652 # spice_ticket => used for spice migration, passed via tunnel/stdin
5653 # network => CIDR of migration network
5654 # type => secure/insecure - tunnel over encrypted connection or plain-text
5655 # nbd_proto_version => int, 0 for TCP, 1 for UNIX
5656 # replicated_volumes => which volids should be re-used with bitmaps for nbd migration
5657 # offline_volumes => new volids of offline migrated disks like tpmstate and cloudinit, not yet
5658 # contained in config
5659 sub vm_start_nolock
{
5660 my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
5662 my $statefile = $params->{statefile
};
5663 my $resume = $params->{resume
};
5665 my $migratedfrom = $migrate_opts->{migratedfrom
};
5666 my $migration_type = $migrate_opts->{type
};
5670 # clean up leftover reboot request files
5671 eval { clear_reboot_request
($vmid); };
5674 if (!$statefile && scalar(keys %{$conf->{pending
}})) {
5675 vmconfig_apply_pending
($vmid, $conf, $storecfg);
5676 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5679 # don't regenerate the ISO if the VM is started as part of a live migration
5680 # this way we can reuse the old ISO with the correct config
5681 if (!$migratedfrom) {
5682 if (PVE
::QemuServer
::Cloudinit
::apply_cloudinit_config
($conf, $vmid)) {
5683 # FIXME: apply_cloudinit_config updates $conf in this case, and it would only drop
5684 # $conf->{cloudinit}, so we could just not do this?
5685 # But we do it above, so for now let's be consistent.
5686 $conf = PVE
::QemuConfig-
>load_config($vmid); # update/reload
5690 # override offline migrated volumes, conf is out of date still
5691 if (my $offline_volumes = $migrate_opts->{offline_volumes
}) {
5692 for my $key (sort keys $offline_volumes->%*) {
5693 my $parsed = parse_drive
($key, $conf->{$key});
5694 $parsed->{file
} = $offline_volumes->{$key};
5695 $conf->{$key} = print_drive
($parsed);
5699 my $defaults = load_defaults
();
5701 # set environment variable useful inside network script
5702 # for remote migration the config is available on the target node!
5703 if (!$migrate_opts->{remote_node
}) {
5704 $ENV{PVE_MIGRATED_FROM
} = $migratedfrom;
5707 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-start', 1);
5709 my $forcemachine = $params->{forcemachine
};
5710 my $forcecpu = $params->{forcecpu
};
5712 # enforce machine and CPU type on suspended vm to ensure HW compatibility
5713 $forcemachine = $conf->{runningmachine
};
5714 $forcecpu = $conf->{runningcpu
};
5715 print "Resuming suspended VM\n";
5718 my ($cmd, $vollist, $spice_port) = config_to_command
($storecfg, $vmid,
5719 $conf, $defaults, $forcemachine, $forcecpu, $params->{'pbs-backing'});
5722 my $get_migration_ip = sub {
5723 my ($nodename) = @_;
5725 return $migration_ip if defined($migration_ip);
5727 my $cidr = $migrate_opts->{network
};
5729 if (!defined($cidr)) {
5730 my $dc_conf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5731 $cidr = $dc_conf->{migration
}->{network
};
5734 if (defined($cidr)) {
5735 my $ips = PVE
::Network
::get_local_ip_from_cidr
($cidr);
5737 die "could not get IP: no address configured on local " .
5738 "node for network '$cidr'\n" if scalar(@$ips) == 0;
5740 die "could not get IP: multiple addresses configured on local " .
5741 "node for network '$cidr'\n" if scalar(@$ips) > 1;
5743 $migration_ip = @$ips[0];
5746 $migration_ip = PVE
::Cluster
::remote_node_ip
($nodename, 1)
5747 if !defined($migration_ip);
5749 return $migration_ip;
5753 if ($statefile eq 'tcp') {
5754 my $migrate = $res->{migrate
} = { proto
=> 'tcp' };
5755 $migrate->{addr
} = "localhost";
5756 my $datacenterconf = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
5757 my $nodename = nodename
();
5759 if (!defined($migration_type)) {
5760 if (defined($datacenterconf->{migration
}->{type
})) {
5761 $migration_type = $datacenterconf->{migration
}->{type
};
5763 $migration_type = 'secure';
5767 if ($migration_type eq 'insecure') {
5768 $migrate->{addr
} = $get_migration_ip->($nodename);
5769 $migrate->{addr
} = "[$migrate->{addr}]" if Net
::IP
::ip_is_ipv6
($migrate->{addr
});
5772 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5773 $migrate->{port
} = PVE
::Tools
::next_migrate_port
($pfamily);
5774 $migrate->{uri
} = "tcp:$migrate->{addr}:$migrate->{port}";
5775 push @$cmd, '-incoming', $migrate->{uri
};
5778 } elsif ($statefile eq 'unix') {
5779 # should be default for secure migrations as a ssh TCP forward
5780 # tunnel is not deterministic reliable ready and fails regurarly
5781 # to set up in time, so use UNIX socket forwards
5782 my $migrate = $res->{migrate
} = { proto
=> 'unix' };
5783 $migrate->{addr
} = "/run/qemu-server/$vmid.migrate";
5784 unlink $migrate->{addr
};
5786 $migrate->{uri
} = "unix:$migrate->{addr}";
5787 push @$cmd, '-incoming', $migrate->{uri
};
5790 } elsif (-e
$statefile) {
5791 push @$cmd, '-loadstate', $statefile;
5793 my $statepath = PVE
::Storage
::path
($storecfg, $statefile);
5794 push @$vollist, $statefile;
5795 push @$cmd, '-loadstate', $statepath;
5797 } elsif ($params->{paused
}) {
5801 my $start_timeout = $params->{timeout
} // config_aware_timeout
($conf, $resume);
5803 my $pci_devices = {}; # host pci devices
5804 for (my $i = 0; $i < $PVE::QemuServer
::PCI
::MAX_HOSTPCI_DEVICES
; $i++) {
5805 my $dev = $conf->{"hostpci$i"} or next;
5806 $pci_devices->{$i} = parse_hostpci
($dev);
5809 # do not reserve pciid for mediated devices, sysfs will error out for duplicate assignment
5810 my $real_pci_devices = [ grep { !(defined($_->{mdev
}) && scalar($_->{pciid
}->@*) == 1) } values $pci_devices->%* ];
5812 # map to a flat list of pci ids
5813 my $pci_id_list = [ map { $_->{id
} } map { $_->{pciid
}->@* } $real_pci_devices->@* ];
5815 # reserve all PCI IDs before actually doing anything with them
5816 PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, $start_timeout);
5820 for my $id (sort keys %$pci_devices) {
5821 my $d = $pci_devices->{$id};
5822 for my $dev ($d->{pciid
}->@*) {
5823 my $info = PVE
::QemuServer
::PCI
::prepare_pci_device
($vmid, $dev->{id
}, $id, $d->{mdev
});
5825 # nvidia grid needs the uuid of the mdev as qemu parameter
5826 if ($d->{mdev
} && !defined($uuid) && $info->{vendor
} eq '10de') {
5827 $uuid = PVE
::QemuServer
::PCI
::generate_mdev_uuid
($vmid, $id);
5831 push @$cmd, '-uuid', $uuid if defined($uuid);
5834 eval { cleanup_pci_devices
($vmid, $conf) };
5839 PVE
::Storage
::activate_volumes
($storecfg, $vollist);
5842 run_command
(['/bin/systemctl', 'stop', "$vmid.scope"], outfunc
=> sub{}, errfunc
=> sub{});
5844 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5845 # timeout should be more than enough here...
5846 PVE
::Systemd
::wait_for_unit_removed
("$vmid.scope", 20);
5848 my $cpuunits = PVE
::CGroup
::clamp_cpu_shares
($conf->{cpuunits
});
5851 timeout
=> $statefile ?
undef : $start_timeout,
5856 # when migrating, prefix QEMU output so other side can pick up any
5857 # errors that might occur and show the user
5858 if ($migratedfrom) {
5859 $run_params{quiet
} = 1;
5860 $run_params{logfunc
} = sub { print "QEMU: $_[0]\n" };
5863 my %systemd_properties = (
5864 Slice
=> 'qemu.slice',
5865 KillMode
=> 'process',
5867 TimeoutStopUSec
=> ULONG_MAX
, # infinity
5870 if (PVE
::CGroup
::cgroup_mode
() == 2) {
5871 $systemd_properties{CPUWeight
} = $cpuunits;
5873 $systemd_properties{CPUShares
} = $cpuunits;
5876 if (my $cpulimit = $conf->{cpulimit
}) {
5877 $systemd_properties{CPUQuota
} = int($cpulimit * 100);
5879 $systemd_properties{timeout
} = 10 if $statefile; # setting up the scope shoul be quick
5881 my $run_qemu = sub {
5882 PVE
::Tools
::run_fork
sub {
5883 PVE
::Systemd
::enter_systemd_scope
($vmid, "Proxmox VE VM $vmid", %systemd_properties);
5886 if (my $tpm = $conf->{tpmstate0
}) {
5887 # start the TPM emulator so QEMU can connect on start
5888 $tpmpid = start_swtpm
($storecfg, $vmid, $tpm, $migratedfrom);
5891 my $exitcode = run_command
($cmd, %run_params);
5894 warn "stopping swtpm instance (pid $tpmpid) due to QEMU startup error\n";
5895 kill 'TERM', $tpmpid;
5897 die "QEMU exited with code $exitcode\n";
5902 if ($conf->{hugepages
}) {
5905 my $hugepages_topology = PVE
::QemuServer
::Memory
::hugepages_topology
($conf);
5906 my $hugepages_host_topology = PVE
::QemuServer
::Memory
::hugepages_host_topology
();
5908 PVE
::QemuServer
::Memory
::hugepages_mount
();
5909 PVE
::QemuServer
::Memory
::hugepages_allocate
($hugepages_topology, $hugepages_host_topology);
5911 eval { $run_qemu->() };
5913 PVE
::QemuServer
::Memory
::hugepages_reset
($hugepages_host_topology)
5914 if !$conf->{keephugepages
};
5918 PVE
::QemuServer
::Memory
::hugepages_pre_deallocate
($hugepages_topology)
5919 if !$conf->{keephugepages
};
5921 eval { PVE
::QemuServer
::Memory
::hugepages_update_locked
($code); };
5924 eval { $run_qemu->() };
5928 # deactivate volumes if start fails
5929 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
5931 eval { cleanup_pci_devices
($vmid, $conf) };
5934 die "start failed: $err";
5937 # re-reserve all PCI IDs now that we can know the actual VM PID
5938 my $pid = PVE
::QemuServer
::Helpers
::vm_running_locally
($vmid);
5939 eval { PVE
::QemuServer
::PCI
::reserve_pci_usage
($pci_id_list, $vmid, undef, $pid) };
5942 if (defined($res->{migrate
})) {
5943 print "migration listens on $res->{migrate}->{uri}\n";
5944 } elsif ($statefile) {
5945 eval { mon_cmd
($vmid, "cont"); };
5949 #start nbd server for storage migration
5950 if (my $nbd = $migrate_opts->{nbd
}) {
5951 my $nbd_protocol_version = $migrate_opts->{nbd_proto_version
} // 0;
5953 my $migrate_storage_uri;
5954 # nbd_protocol_version > 0 for unix socket support
5955 if ($nbd_protocol_version > 0 && ($migration_type eq 'secure' || $migration_type eq 'websocket')) {
5956 my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
5957 mon_cmd
($vmid, "nbd-server-start", addr
=> { type
=> 'unix', data
=> { path
=> $socket_path } } );
5958 $migrate_storage_uri = "nbd:unix:$socket_path";
5959 $res->{migrate
}->{unix_sockets
} = [$socket_path];
5961 my $nodename = nodename
();
5962 my $localip = $get_migration_ip->($nodename);
5963 my $pfamily = PVE
::Tools
::get_host_address_family
($nodename);
5964 my $storage_migrate_port = PVE
::Tools
::next_migrate_port
($pfamily);
5966 mon_cmd
($vmid, "nbd-server-start", addr
=> {
5969 host
=> "${localip}",
5970 port
=> "${storage_migrate_port}",
5973 $localip = "[$localip]" if Net
::IP
::ip_is_ipv6
($localip);
5974 $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
5977 my $block_info = mon_cmd
($vmid, "query-block");
5978 $block_info = { map { $_->{device
} => $_ } $block_info->@* };
5980 foreach my $opt (sort keys %$nbd) {
5981 my $drivestr = $nbd->{$opt}->{drivestr
};
5982 my $volid = $nbd->{$opt}->{volid
};
5984 my $block_node = $block_info->{"drive-$opt"}->{inserted
}->{'node-name'};
5990 'node-name' => $block_node,
5991 writable
=> JSON
::true
,
5993 name
=> "drive-$opt", # NBD export name
5996 my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
5997 print "storage migration listens on $nbd_uri volume:$drivestr\n";
5998 print "re-using replicated volume: $opt - $volid\n"
5999 if $nbd->{$opt}->{replicated
};
6001 $res->{drives
}->{$opt} = $nbd->{$opt};
6002 $res->{drives
}->{$opt}->{nbd_uri
} = $nbd_uri;
6006 if ($migratedfrom) {
6008 set_migration_caps
($vmid);
6013 print "spice listens on port $spice_port\n";
6014 $res->{spice_port
} = $spice_port;
6015 if ($migrate_opts->{spice_ticket
}) {
6016 mon_cmd
($vmid, "set_password", protocol
=> 'spice', password
=>
6017 $migrate_opts->{spice_ticket
});
6018 mon_cmd
($vmid, "expire_password", protocol
=> 'spice', time => "+30");
6023 mon_cmd
($vmid, "balloon", value
=> $conf->{balloon
}*1024*1024)
6024 if !$statefile && $conf->{balloon
};
6026 foreach my $opt (keys %$conf) {
6027 next if $opt !~ m/^net\d+$/;
6028 my $nicconf = parse_net
($conf->{$opt});
6029 qemu_set_link_status
($vmid, $opt, 0) if $nicconf->{link_down
};
6031 add_nets_bridge_fdb
($conf, $vmid);
6034 if (!defined($conf->{balloon
}) || $conf->{balloon
}) {
6038 path
=> "machine/peripheral/balloon0",
6039 property
=> "guest-stats-polling-interval",
6045 print "Resumed VM, removing state\n";
6046 if (my $vmstate = $conf->{vmstate
}) {
6047 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6048 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6050 delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
6051 PVE
::QemuConfig-
>write_config($vmid, $conf);
6054 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'post-start');
6059 sub vm_commandline
{
6060 my ($storecfg, $vmid, $snapname) = @_;
6062 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6064 my ($forcemachine, $forcecpu);
6066 my $snapshot = $conf->{snapshots
}->{$snapname};
6067 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
6069 # check for machine or CPU overrides in snapshot
6070 $forcemachine = $snapshot->{runningmachine
};
6071 $forcecpu = $snapshot->{runningcpu
};
6073 $snapshot->{digest
} = $conf->{digest
}; # keep file digest for API
6078 my $defaults = load_defaults
();
6080 my $cmd = config_to_command
($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
6082 return PVE
::Tools
::cmd2string
($cmd);
6086 my ($vmid, $skiplock) = @_;
6088 PVE
::QemuConfig-
>lock_config($vmid, sub {
6090 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6092 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6094 mon_cmd
($vmid, "system_reset");
6098 sub get_vm_volumes
{
6102 foreach_volid
($conf, sub {
6103 my ($volid, $attr) = @_;
6105 return if $volid =~ m
|^/|;
6107 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
6110 push @$vollist, $volid;
6116 sub cleanup_pci_devices
{
6117 my ($vmid, $conf) = @_;
6119 foreach my $key (keys %$conf) {
6120 next if $key !~ m/^hostpci(\d+)$/;
6121 my $hostpciindex = $1;
6122 my $uuid = PVE
::SysFSTools
::generate_mdev_uuid
($vmid, $hostpciindex);
6123 my $d = parse_hostpci
($conf->{$key});
6125 # NOTE: avoid PVE::SysFSTools::pci_cleanup_mdev_device as it requires PCI ID and we
6126 # don't want to break ABI just for this two liner
6127 my $dev_sysfs_dir = "/sys/bus/mdev/devices/$uuid";
6128 PVE
::SysFSTools
::file_write
("$dev_sysfs_dir/remove", "1") if -e
$dev_sysfs_dir;
6131 PVE
::QemuServer
::PCI
::remove_pci_reservation
($vmid);
6134 sub vm_stop_cleanup
{
6135 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
6140 my $vollist = get_vm_volumes
($conf);
6141 PVE
::Storage
::deactivate_volumes
($storecfg, $vollist);
6143 if (my $tpmdrive = $conf->{tpmstate0
}) {
6144 my $tpm = parse_drive
("tpmstate0", $tpmdrive);
6145 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($tpm->{file
}, 1);
6147 PVE
::Storage
::unmap_volume
($storecfg, $tpm->{file
});
6152 foreach my $ext (qw(mon qmp pid vnc qga)) {
6153 unlink "/var/run/qemu-server/${vmid}.$ext";
6156 if ($conf->{ivshmem
}) {
6157 my $ivshmem = parse_property_string
($ivshmem_fmt, $conf->{ivshmem
});
6158 # just delete it for now, VMs which have this already open do not
6159 # are affected, but new VMs will get a separated one. If this
6160 # becomes an issue we either add some sort of ref-counting or just
6161 # add a "don't delete on stop" flag to the ivshmem format.
6162 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name
} // $vmid);
6165 cleanup_pci_devices
($vmid, $conf);
6167 vmconfig_apply_pending
($vmid, $conf, $storecfg) if $apply_pending_changes;
6169 warn $@ if $@; # avoid errors - just warn
6172 # call only in locked context
6174 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
6176 my $pid = check_running
($vmid, $nocheck);
6181 $conf = PVE
::QemuConfig-
>load_config($vmid);
6182 PVE
::QemuConfig-
>check_lock($conf) if !$skiplock;
6183 if (!defined($timeout) && $shutdown && $conf->{startup
}) {
6184 my $opts = PVE
::JSONSchema
::pve_parse_startup_order
($conf->{startup
});
6185 $timeout = $opts->{down
} if $opts->{down
};
6187 PVE
::GuestHelpers
::exec_hookscript
($conf, $vmid, 'pre-stop');
6192 if (defined($conf) && get_qga_key
($conf, 'enabled')) {
6193 mon_cmd
($vmid, "guest-shutdown", timeout
=> $timeout);
6195 mon_cmd
($vmid, "system_powerdown");
6198 mon_cmd
($vmid, "quit");
6204 $timeout = 60 if !defined($timeout);
6207 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6212 if ($count >= $timeout) {
6214 warn "VM still running - terminating now with SIGTERM\n";
6217 die "VM quit/powerdown failed - got timeout\n";
6220 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6224 if (!check_running
($vmid, $nocheck)) {
6225 warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
6229 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
6232 die "VM quit/powerdown failed\n";
6240 while (($count < $timeout) && check_running
($vmid, $nocheck)) {
6245 if ($count >= $timeout) {
6246 warn "VM still running - terminating now with SIGKILL\n";
6251 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
6254 # Note: use $nocheck to skip tests if VM configuration file exists.
6255 # We need that when migration VMs to other nodes (files already moved)
6256 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
6258 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
6260 $force = 1 if !defined($force) && !$shutdown;
6263 my $pid = check_running
($vmid, $nocheck, $migratedfrom);
6264 kill 15, $pid if $pid;
6265 my $conf = PVE
::QemuConfig-
>load_config($vmid, $migratedfrom);
6266 vm_stop_cleanup
($storecfg, $vmid, $conf, $keepActive, 0);
6270 PVE
::QemuConfig-
>lock_config($vmid, sub {
6271 _do_vm_stop
($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
6276 my ($vmid, $timeout) = @_;
6278 PVE
::QemuConfig-
>lock_config($vmid, sub {
6281 # only reboot if running, as qmeventd starts it again on a stop event
6282 return if !check_running
($vmid);
6284 create_reboot_request
($vmid);
6286 my $storecfg = PVE
::Storage
::config
();
6287 _do_vm_stop
($storecfg, $vmid, undef, undef, $timeout, 1);
6291 # avoid that the next normal shutdown will be confused for a reboot
6292 clear_reboot_request
($vmid);
6298 # note: if using the statestorage parameter, the caller has to check privileges
6300 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
6307 PVE
::QemuConfig-
>lock_config($vmid, sub {
6309 $conf = PVE
::QemuConfig-
>load_config($vmid);
6311 my $is_backing_up = PVE
::QemuConfig-
>has_lock($conf, 'backup');
6312 PVE
::QemuConfig-
>check_lock($conf)
6313 if !($skiplock || $is_backing_up);
6315 die "cannot suspend to disk during backup\n"
6316 if $is_backing_up && $includestate;
6318 if ($includestate) {
6319 $conf->{lock} = 'suspending';
6320 my $date = strftime
("%Y-%m-%d", localtime(time()));
6321 $storecfg = PVE
::Storage
::config
();
6322 if (!$statestorage) {
6323 $statestorage = find_vmstate_storage
($conf, $storecfg);
6324 # check permissions for the storage
6325 my $rpcenv = PVE
::RPCEnvironment
::get
();
6326 if ($rpcenv->{type
} ne 'cli') {
6327 my $authuser = $rpcenv->get_user();
6328 $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
6333 $vmstate = PVE
::QemuConfig-
>__snapshot_save_vmstate(
6334 $vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
6335 $path = PVE
::Storage
::path
($storecfg, $vmstate);
6336 PVE
::QemuConfig-
>write_config($vmid, $conf);
6338 mon_cmd
($vmid, "stop");
6342 if ($includestate) {
6344 PVE
::Storage
::activate_volumes
($storecfg, [$vmstate]);
6347 set_migration_caps
($vmid, 1);
6348 mon_cmd
($vmid, "savevm-start", statefile
=> $path);
6350 my $state = mon_cmd
($vmid, "query-savevm");
6351 if (!$state->{status
}) {
6352 die "savevm not active\n";
6353 } elsif ($state->{status
} eq 'active') {
6356 } elsif ($state->{status
} eq 'completed') {
6357 print "State saved, quitting\n";
6359 } elsif ($state->{status
} eq 'failed' && $state->{error
}) {
6360 die "query-savevm failed with error '$state->{error}'\n"
6362 die "query-savevm returned status '$state->{status}'\n";
6368 PVE
::QemuConfig-
>lock_config($vmid, sub {
6369 $conf = PVE
::QemuConfig-
>load_config($vmid);
6371 # cleanup, but leave suspending lock, to indicate something went wrong
6373 mon_cmd
($vmid, "savevm-end");
6374 PVE
::Storage
::deactivate_volumes
($storecfg, [$vmstate]);
6375 PVE
::Storage
::vdisk_free
($storecfg, $vmstate);
6376 delete $conf->@{qw(vmstate runningmachine runningcpu)};
6377 PVE
::QemuConfig-
>write_config($vmid, $conf);
6383 die "lock changed unexpectedly\n"
6384 if !PVE
::QemuConfig-
>has_lock($conf, 'suspending');
6386 mon_cmd
($vmid, "quit");
6387 $conf->{lock} = 'suspended';
6388 PVE
::QemuConfig-
>write_config($vmid, $conf);
6393 # $nocheck is set when called as part of a migration - in this context the
6394 # location of the config file (source or target node) is not deterministic,
6395 # since migration cannot wait for pmxcfs to process the rename
6397 my ($vmid, $skiplock, $nocheck) = @_;
6399 PVE
::QemuConfig-
>lock_config($vmid, sub {
6400 my $res = mon_cmd
($vmid, 'query-status');
6401 my $resume_cmd = 'cont';
6405 $conf = eval { PVE
::QemuConfig-
>load_config($vmid) }; # try on target node
6407 my $vmlist = PVE
::Cluster
::get_vmlist
();
6408 if (exists($vmlist->{ids
}->{$vmid})) {
6409 my $node = $vmlist->{ids
}->{$vmid}->{node
};
6410 $conf = eval { PVE
::QemuConfig-
>load_config($vmid, $node) }; # try on source node
6413 PVE
::Cluster
::cfs_update
(); # vmlist was wrong, invalidate cache
6414 $conf = PVE
::QemuConfig-
>load_config($vmid); # last try on target node again
6418 $conf = PVE
::QemuConfig-
>load_config($vmid);
6421 if ($res->{status
}) {
6422 return if $res->{status
} eq 'running'; # job done, go home
6423 $resume_cmd = 'system_wakeup' if $res->{status
} eq 'suspended';
6424 $reset = 1 if $res->{status
} eq 'shutdown';
6428 PVE
::QemuConfig-
>check_lock($conf)
6429 if !($skiplock || PVE
::QemuConfig-
>has_lock($conf, 'backup'));
6433 # required if a VM shuts down during a backup and we get a resume
6434 # request before the backup finishes for example
6435 mon_cmd
($vmid, "system_reset");
6438 add_nets_bridge_fdb
($conf, $vmid) if $resume_cmd eq 'cont';
6440 mon_cmd
($vmid, $resume_cmd);
6445 my ($vmid, $skiplock, $key) = @_;
6447 PVE
::QemuConfig-
>lock_config($vmid, sub {
6449 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6451 # there is no qmp command, so we use the human monitor command
6452 my $res = PVE
::QemuServer
::Monitor
::hmp_cmd
($vmid, "sendkey $key");
6453 die $res if $res ne '';
6457 # vzdump restore implementaion
6459 sub tar_archive_read_firstfile
{
6460 my $archive = shift;
6462 die "ERROR: file '$archive' does not exist\n" if ! -f
$archive;
6464 # try to detect archive type first
6465 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6466 die "unable to open file '$archive'\n";
6467 my $firstfile = <$fh>;
6471 die "ERROR: archive contaions no data\n" if !$firstfile;
6477 sub tar_restore_cleanup
{
6478 my ($storecfg, $statfile) = @_;
6480 print STDERR
"starting cleanup\n";
6482 if (my $fd = IO
::File-
>new($statfile, "r")) {
6483 while (defined(my $line = <$fd>)) {
6484 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6487 if ($volid =~ m
|^/|) {
6488 unlink $volid || die 'unlink failed\n';
6490 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6492 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6494 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6496 print STDERR
"unable to parse line in statfile - $line";
6503 sub restore_file_archive
{
6504 my ($archive, $vmid, $user, $opts) = @_;
6506 return restore_vma_archive
($archive, $vmid, $user, $opts)
6509 my $info = PVE
::Storage
::archive_info
($archive);
6510 my $format = $opts->{format
} // $info->{format
};
6511 my $comp = $info->{compression
};
6513 # try to detect archive format
6514 if ($format eq 'tar') {
6515 return restore_tar_archive
($archive, $vmid, $user, $opts);
6517 return restore_vma_archive
($archive, $vmid, $user, $opts, $comp);
6521 # hepler to remove disks that will not be used after restore
6522 my $restore_cleanup_oldconf = sub {
6523 my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
6525 my $kept_disks = {};
6527 PVE
::QemuConfig-
>foreach_volume($oldconf, sub {
6528 my ($ds, $drive) = @_;
6530 return if drive_is_cdrom
($drive, 1);
6532 my $volid = $drive->{file
};
6533 return if !$volid || $volid =~ m
|^/|;
6535 my ($path, $owner) = PVE
::Storage
::path
($storecfg, $volid);
6536 return if !$path || !$owner || ($owner != $vmid);
6538 # Note: only delete disk we want to restore
6539 # other volumes will become unused
6540 if ($virtdev_hash->{$ds}) {
6541 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
6546 $kept_disks->{$volid} = 1;
6550 # after the restore we have no snapshots anymore
6551 for my $snapname (keys $oldconf->{snapshots
}->%*) {
6552 my $snap = $oldconf->{snapshots
}->{$snapname};
6553 if ($snap->{vmstate
}) {
6554 eval { PVE
::Storage
::vdisk_free
($storecfg, $snap->{vmstate
}); };
6560 for my $volid (keys $kept_disks->%*) {
6561 eval { PVE
::Storage
::volume_snapshot_delete
($storecfg, $volid, $snapname); };
6567 # Helper to parse vzdump backup device hints
6569 # $rpcenv: Environment, used to ckeck storage permissions
6570 # $user: User ID, to check storage permissions
6571 # $storecfg: Storage configuration
6572 # $fh: the file handle for reading the configuration
6573 # $devinfo: should contain device sizes for all backu-up'ed devices
6574 # $options: backup options (pool, default storage)
6576 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
6577 my $parse_backup_hints = sub {
6578 my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
6580 my $check_storage = sub { # assert if an image can be allocate
6581 my ($storeid, $scfg) = @_;
6582 die "Content type 'images' is not available on storage '$storeid'\n"
6583 if !$scfg->{content
}->{images
};
6584 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace'])
6585 if $user ne 'root@pam';
6588 my $virtdev_hash = {};
6589 while (defined(my $line = <$fh>)) {
6590 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6591 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6592 die "archive does not contain data for drive '$virtdev'\n"
6593 if !$devinfo->{$devname};
6595 if (defined($options->{storage
})) {
6596 $storeid = $options->{storage
} || 'local';
6597 } elsif (!$storeid) {
6600 $format = 'raw' if !$format;
6601 $devinfo->{$devname}->{devname
} = $devname;
6602 $devinfo->{$devname}->{virtdev
} = $virtdev;
6603 $devinfo->{$devname}->{format
} = $format;
6604 $devinfo->{$devname}->{storeid
} = $storeid;
6606 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6607 $check_storage->($storeid, $scfg); # permission and content type check
6609 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6610 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6612 my $drive = parse_drive
($virtdev, $2);
6614 if (drive_is_cloudinit
($drive)) {
6615 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
6616 $storeid = $options->{storage
} if defined ($options->{storage
});
6617 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6618 my $format = qemu_img_format
($scfg, $volname); # has 'raw' fallback
6620 $check_storage->($storeid, $scfg); # permission and content type check
6622 $virtdev_hash->{$virtdev} = {
6624 storeid
=> $storeid,
6625 size
=> PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
,
6632 return $virtdev_hash;
6635 # Helper to allocate and activate all volumes required for a restore
6637 # $storecfg: Storage configuration
6638 # $virtdev_hash: as returned by parse_backup_hints()
6640 # Returns: { $virtdev => $volid }
6641 my $restore_allocate_devices = sub {
6642 my ($storecfg, $virtdev_hash, $vmid) = @_;
6645 foreach my $virtdev (sort keys %$virtdev_hash) {
6646 my $d = $virtdev_hash->{$virtdev};
6647 my $alloc_size = int(($d->{size
} + 1024 - 1)/1024);
6648 my $storeid = $d->{storeid
};
6649 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6651 # test if requested format is supported
6652 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
6653 my $supported = grep { $_ eq $d->{format
} } @$validFormats;
6654 $d->{format
} = $defFormat if !$supported;
6657 if ($d->{is_cloudinit
}) {
6658 $name = "vm-$vmid-cloudinit";
6659 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6660 if ($scfg->{path
}) {
6661 $name .= ".$d->{format}";
6665 my $volid = PVE
::Storage
::vdisk_alloc
(
6666 $storecfg, $storeid, $vmid, $d->{format
}, $name, $alloc_size);
6668 print STDERR
"new volume ID is '$volid'\n";
6669 $d->{volid
} = $volid;
6671 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
6673 $map->{$virtdev} = $volid;
6679 sub restore_update_config_line
{
6680 my ($cookie, $map, $line, $unique) = @_;
6682 return '' if $line =~ m/^\#qmdump\#/;
6683 return '' if $line =~ m/^\#vzdump\#/;
6684 return '' if $line =~ m/^lock:/;
6685 return '' if $line =~ m/^unused\d+:/;
6686 return '' if $line =~ m/^parent:/;
6690 my $dc = PVE
::Cluster
::cfs_read_file
('datacenter.cfg');
6691 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6692 # try to convert old 1.X settings
6693 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6694 foreach my $devconfig (PVE
::Tools
::split_list
($ethcfg)) {
6695 my ($model, $macaddr) = split(/\=/, $devconfig);
6696 $macaddr = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if !$macaddr || $unique;
6699 bridge
=> "vmbr$ind",
6700 macaddr
=> $macaddr,
6702 my $netstr = print_net
($net);
6704 $res .= "net$cookie->{netcount}: $netstr\n";
6705 $cookie->{netcount
}++;
6707 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6708 my ($id, $netstr) = ($1, $2);
6709 my $net = parse_net
($netstr);
6710 $net->{macaddr
} = PVE
::Tools
::random_ether_addr
($dc->{mac_prefix
}) if $net->{macaddr
};
6711 $netstr = print_net
($net);
6712 $res .= "$id: $netstr\n";
6713 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk|tpmstate)\d+):\s*(\S+)\s*$/) {
6716 my $di = parse_drive
($virtdev, $value);
6717 if (defined($di->{backup
}) && !$di->{backup
}) {
6719 } elsif ($map->{$virtdev}) {
6720 delete $di->{format
}; # format can change on restore
6721 $di->{file
} = $map->{$virtdev};
6722 $value = print_drive
($di);
6723 $res .= "$virtdev: $value\n";
6727 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6729 if ($vmgenid ne '0') {
6730 # always generate a new vmgenid if there was a valid one setup
6731 $vmgenid = generate_uuid
();
6733 $res .= "vmgenid: $vmgenid\n";
6734 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6735 my ($uuid, $uuid_str);
6736 UUID
::generate
($uuid);
6737 UUID
::unparse
($uuid, $uuid_str);
6738 my $smbios1 = parse_smbios1
($2);
6739 $smbios1->{uuid
} = $uuid_str;
6740 $res .= $1.print_smbios1
($smbios1)."\n";
6748 my $restore_deactivate_volumes = sub {
6749 my ($storecfg, $virtdev_hash) = @_;
6752 for my $dev (values $virtdev_hash->%*) {
6753 push $vollist->@*, $dev->{volid
} if $dev->{volid
};
6756 eval { PVE
::Storage
::deactivate_volumes
($storecfg, $vollist); };
6757 print STDERR
$@ if $@;
6760 my $restore_destroy_volumes = sub {
6761 my ($storecfg, $virtdev_hash) = @_;
6763 for my $dev (values $virtdev_hash->%*) {
6764 my $volid = $dev->{volid
} or next;
6766 PVE
::Storage
::vdisk_free
($storecfg, $volid);
6767 print STDERR
"temporary volume '$volid' sucessfuly removed\n";
6769 print STDERR
"unable to cleanup '$volid' - $@" if $@;
6773 my $restore_merge_config = sub {
6774 my ($filename, $backup_conf_raw, $override_conf) = @_;
6776 my $backup_conf = parse_vm_config
($filename, $backup_conf_raw);
6777 for my $key (keys $override_conf->%*) {
6778 $backup_conf->{$key} = $override_conf->{$key};
6781 return $backup_conf;
6785 my ($cfg, $vmid) = @_;
6787 my $info = PVE
::Storage
::vdisk_list
($cfg, undef, $vmid, undef, 'images');
6789 my $volid_hash = {};
6790 foreach my $storeid (keys %$info) {
6791 foreach my $item (@{$info->{$storeid}}) {
6792 next if !($item->{volid
} && $item->{size
});
6793 $item->{path
} = PVE
::Storage
::path
($cfg, $item->{volid
});
6794 $volid_hash->{$item->{volid
}} = $item;
6801 sub update_disk_config
{
6802 my ($vmid, $conf, $volid_hash) = @_;
6805 my $prefix = "VM $vmid";
6807 # used and unused disks
6808 my $referenced = {};
6810 # Note: it is allowed to define multiple storages with same path (alias), so
6811 # we need to check both 'volid' and real 'path' (two different volid can point
6812 # to the same path).
6814 my $referencedpath = {};
6817 PVE
::QemuConfig-
>foreach_volume($conf, sub {
6818 my ($opt, $drive) = @_;
6820 my $volid = $drive->{file
};
6822 my $volume = $volid_hash->{$volid};
6824 # mark volid as "in-use" for next step
6825 $referenced->{$volid} = 1;
6826 if ($volume && (my $path = $volume->{path
})) {
6827 $referencedpath->{$path} = 1;
6830 return if drive_is_cdrom
($drive);
6833 my ($updated, $msg) = PVE
::QemuServer
::Drive
::update_disksize
($drive, $volume->{size
});
6834 if (defined($updated)) {
6836 $conf->{$opt} = print_drive
($updated);
6837 print "$prefix ($opt): $msg\n";
6841 # remove 'unusedX' entry if volume is used
6842 PVE
::QemuConfig-
>foreach_unused_volume($conf, sub {
6843 my ($opt, $drive) = @_;
6845 my $volid = $drive->{file
};
6849 $path = $volid_hash->{$volid}->{path
} if $volid_hash->{$volid};
6850 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6851 print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
6853 delete $conf->{$opt};
6856 $referenced->{$volid} = 1;
6857 $referencedpath->{$path} = 1 if $path;
6860 foreach my $volid (sort keys %$volid_hash) {
6861 next if $volid =~ m/vm-$vmid-state-/;
6862 next if $referenced->{$volid};
6863 my $path = $volid_hash->{$volid}->{path
};
6864 next if !$path; # just to be sure
6865 next if $referencedpath->{$path};
6867 my $key = PVE
::QemuConfig-
>add_unused_volume($conf, $volid);
6868 print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
6869 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6876 my ($vmid, $nolock, $dryrun) = @_;
6878 my $cfg = PVE
::Storage
::config
();
6880 print "rescan volumes...\n";
6881 my $volid_hash = scan_volids
($cfg, $vmid);
6883 my $updatefn = sub {
6886 my $conf = PVE
::QemuConfig-
>load_config($vmid);
6888 PVE
::QemuConfig-
>check_lock($conf);
6891 foreach my $volid (keys %$volid_hash) {
6892 my $info = $volid_hash->{$volid};
6893 $vm_volids->{$volid} = $info if $info->{vmid
} && $info->{vmid
} == $vmid;
6896 my $changes = update_disk_config
($vmid, $conf, $vm_volids);
6898 PVE
::QemuConfig-
>write_config($vmid, $conf) if $changes && !$dryrun;
6901 if (defined($vmid)) {
6905 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6908 my $vmlist = config_list
();
6909 foreach my $vmid (keys %$vmlist) {
6913 PVE
::QemuConfig-
>lock_config($vmid, $updatefn, $vmid);
6919 sub restore_proxmox_backup_archive
{
6920 my ($archive, $vmid, $user, $options) = @_;
6922 my $storecfg = PVE
::Storage
::config
();
6924 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($archive);
6925 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
6927 my $fingerprint = $scfg->{fingerprint
};
6928 my $keyfile = PVE
::Storage
::PBSPlugin
::pbs_encryption_key_file_name
($storecfg, $storeid);
6930 my $repo = PVE
::PBSClient
::get_repository
($scfg);
6931 my $namespace = $scfg->{namespace
};
6933 # This is only used for `pbs-restore` and the QEMU PBS driver (live-restore)
6934 my $password = PVE
::Storage
::PBSPlugin
::pbs_get_password
($scfg, $storeid);
6935 local $ENV{PBS_PASSWORD
} = $password;
6936 local $ENV{PBS_FINGERPRINT
} = $fingerprint if defined($fingerprint);
6938 my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6939 PVE
::Storage
::parse_volname
($storecfg, $archive);
6941 die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
6943 die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
6945 my $tmpdir = "/var/tmp/vzdumptmp$$";
6949 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
6950 # disable interrupts (always do cleanups)
6954 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
6956 # Note: $oldconf is undef if VM does not exists
6957 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
6958 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
6959 my $new_conf_raw = '';
6961 my $rpcenv = PVE
::RPCEnvironment
::get
();
6962 my $devinfo = {}; # info about drives included in backup
6963 my $virtdev_hash = {}; # info about allocated drives
6971 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
6973 my $cfgfn = "$tmpdir/qemu-server.conf";
6974 my $firewall_config_fn = "$tmpdir/fw.conf";
6975 my $index_fn = "$tmpdir/index.json";
6977 my $cmd = "restore";
6979 my $param = [$pbs_backup_name, "index.json", $index_fn];
6980 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
6981 my $index = PVE
::Tools
::file_get_contents
($index_fn);
6982 $index = decode_json
($index);
6984 foreach my $info (@{$index->{files
}}) {
6985 if ($info->{filename
} =~ m/^(drive-\S+).img.fidx$/) {
6987 if ($info->{size
} =~ m/^(\d+)$/) { # untaint size
6988 $devinfo->{$devname}->{size
} = $1;
6990 die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
6995 my $is_qemu_server_backup = scalar(
6996 grep { $_->{filename
} eq 'qemu-server.conf.blob' } @{$index->{files
}}
6998 if (!$is_qemu_server_backup) {
6999 die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
7001 my $has_firewall_config = scalar(grep { $_->{filename
} eq 'fw.conf.blob' } @{$index->{files
}});
7003 $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
7004 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7006 if ($has_firewall_config) {
7007 $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
7008 PVE
::Storage
::PBSPlugin
::run_raw_client_cmd
($scfg, $storeid, $cmd, $param);
7010 my $pve_firewall_dir = '/etc/pve/firewall';
7011 mkdir $pve_firewall_dir; # make sure the dir exists
7012 PVE
::Tools
::file_copy
($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
7015 my $fh = IO
::File-
>new($cfgfn, "r") ||
7016 die "unable to read qemu-server.conf - $!\n";
7018 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
7020 # fixme: rate limit?
7022 # create empty/temp config
7023 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\nlock: create");
7025 $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
7028 my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
7030 foreach my $virtdev (sort keys %$virtdev_hash) {
7031 my $d = $virtdev_hash->{$virtdev};
7032 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7034 # this fails if storage is unavailable
7035 my $volid = $d->{volid
};
7036 my $path = PVE
::Storage
::path
($storecfg, $volid);
7038 # for live-restore we only want to preload the efidisk and TPM state
7039 next if $options->{live
} && $virtdev ne 'efidisk0' && $virtdev ne 'tpmstate0';
7042 if (defined(my $ns = $scfg->{namespace
})) {
7043 @ns_arg = ('--ns', $ns);
7046 my $pbs_restore_cmd = [
7047 '/usr/bin/pbs-restore',
7048 '--repository', $repo,
7051 "$d->{devname}.img.fidx",
7056 push @$pbs_restore_cmd, '--format', $d->{format
} if $d->{format
};
7057 push @$pbs_restore_cmd, '--keyfile', $keyfile if -e
$keyfile;
7059 if (PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $volid)) {
7060 push @$pbs_restore_cmd, '--skip-zero';
7063 my $dbg_cmdstring = PVE
::Tools
::cmd2string
($pbs_restore_cmd);
7064 print "restore proxmox backup image: $dbg_cmdstring\n";
7065 run_command
($pbs_restore_cmd);
7068 $fh->seek(0, 0) || die "seek failed - $!\n";
7070 my $cookie = { netcount
=> 0 };
7071 while (defined(my $line = <$fh>)) {
7072 $new_conf_raw .= restore_update_config_line
(
7084 if ($err || !$options->{live
}) {
7085 $restore_deactivate_volumes->($storecfg, $virtdev_hash);
7091 $restore_destroy_volumes->($storecfg, $virtdev_hash);
7095 if ($options->{live
}) {
7096 # keep lock during live-restore
7097 $new_conf_raw .= "\nlock: create";
7100 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $options->{override_conf
});
7101 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7103 eval { rescan
($vmid, 1); };
7106 PVE
::AccessControl
::add_vm_to_pool
($vmid, $options->{pool
}) if $options->{pool
};
7108 if ($options->{live
}) {
7114 local $SIG{PIPE
} = sub { die "got signal ($!) - abort\n"; };
7116 my $conf = PVE
::QemuConfig-
>load_config($vmid);
7117 die "cannot do live-restore for template\n" if PVE
::QemuConfig-
>is_template($conf);
7119 # these special drives are already restored before start
7120 delete $devinfo->{'drive-efidisk0'};
7121 delete $devinfo->{'drive-tpmstate0-backup'};
7125 keyfile
=> $keyfile,
7126 snapshot
=> $pbs_backup_name,
7127 namespace
=> $namespace,
7129 pbs_live_restore
($vmid, $conf, $storecfg, $devinfo, $pbs_opts);
7131 PVE
::QemuConfig-
>remove_lock($vmid, "create");
7135 sub pbs_live_restore
{
7136 my ($vmid, $conf, $storecfg, $restored_disks, $opts) = @_;
7138 print "starting VM for live-restore\n";
7139 print "repository: '$opts->{repo}', snapshot: '$opts->{snapshot}'\n";
7141 my $pbs_backing = {};
7142 for my $ds (keys %$restored_disks) {
7143 $ds =~ m/^drive-(.*)$/;
7145 $pbs_backing->{$confname} = {
7146 repository
=> $opts->{repo
},
7147 snapshot
=> $opts->{snapshot
},
7148 archive
=> "$ds.img.fidx",
7150 $pbs_backing->{$confname}->{keyfile
} = $opts->{keyfile
} if -e
$opts->{keyfile
};
7151 $pbs_backing->{$confname}->{namespace
} = $opts->{namespace
} if defined($opts->{namespace
});
7153 my $drive = parse_drive
($confname, $conf->{$confname});
7154 print "restoring '$ds' to '$drive->{file}'\n";
7157 my $drives_streamed = 0;
7159 # make sure HA doesn't interrupt our restore by stopping the VM
7160 if (PVE
::HA
::Config
::vm_is_ha_managed
($vmid)) {
7161 run_command
(['ha-manager', 'set', "vm:$vmid", '--state', 'started']);
7164 # start VM with backing chain pointing to PBS backup, environment vars for PBS driver
7165 # in QEMU (PBS_PASSWORD and PBS_FINGERPRINT) are already set by our caller
7166 vm_start_nolock
($storecfg, $vmid, $conf, {paused
=> 1, 'pbs-backing' => $pbs_backing}, {});
7168 my $qmeventd_fd = register_qmeventd_handle
($vmid);
7170 # begin streaming, i.e. data copy from PBS to target disk for every vol,
7171 # this will effectively collapse the backing image chain consisting of
7172 # [target <- alloc-track -> PBS snapshot] to just [target] (alloc-track
7173 # removes itself once all backing images vanish with 'auto-remove=on')
7175 for my $ds (sort keys %$restored_disks) {
7176 my $job_id = "restore-$ds";
7177 mon_cmd
($vmid, 'block-stream',
7178 'job-id' => $job_id,
7181 $jobs->{$job_id} = {};
7184 mon_cmd
($vmid, 'cont');
7185 qemu_drive_mirror_monitor
($vmid, undef, $jobs, 'auto', 0, 'stream');
7187 print "restore-drive jobs finished successfully, removing all tracking block devices"
7188 ." to disconnect from Proxmox Backup Server\n";
7190 for my $ds (sort keys %$restored_disks) {
7191 mon_cmd
($vmid, 'blockdev-del', 'node-name' => "$ds-pbs");
7194 close($qmeventd_fd);
7200 warn "An error occurred during live-restore: $err\n";
7201 _do_vm_stop
($storecfg, $vmid, 1, 1, 10, 0, 1);
7202 die "live-restore failed\n";
7206 sub restore_vma_archive
{
7207 my ($archive, $vmid, $user, $opts, $comp) = @_;
7209 my $readfrom = $archive;
7211 my $cfg = PVE
::Storage
::config
();
7213 my $bwlimit = $opts->{bwlimit
};
7215 my $dbg_cmdstring = '';
7216 my $add_pipe = sub {
7218 push @$commands, $cmd;
7219 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
7220 $dbg_cmdstring .= PVE
::Tools
::cmd2string
($cmd);
7225 if ($archive eq '-') {
7228 # If we use a backup from a PVE defined storage we also consider that
7229 # storage's rate limit:
7230 my (undef, $volid) = PVE
::Storage
::path_to_volume_id
($cfg, $archive);
7231 if (defined($volid)) {
7232 my ($sid, undef) = PVE
::Storage
::parse_volume_id
($volid);
7233 my $readlimit = PVE
::Storage
::get_bandwidth_limit
('restore', [$sid], $bwlimit);
7235 print STDERR
"applying read rate limit: $readlimit\n";
7236 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
7237 $add_pipe->($cstream);
7243 my $info = PVE
::Storage
::decompressor_info
('vma', $comp);
7244 my $cmd = $info->{decompressor
};
7245 push @$cmd, $readfrom;
7249 my $tmpdir = "/var/tmp/vzdumptmp$$";
7252 # disable interrupts (always do cleanups)
7256 local $SIG{HUP
} = sub { warn "got interrupt - ignored\n"; };
7258 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
7259 POSIX
::mkfifo
($mapfifo, 0600);
7261 my $openfifo = sub { open($fifofh, '>', $mapfifo) or die $! };
7263 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
7268 my $devinfo = {}; # info about drives included in backup
7269 my $virtdev_hash = {}; # info about allocated drives
7271 my $rpcenv = PVE
::RPCEnvironment
::get
();
7273 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7275 # Note: $oldconf is undef if VM does not exist
7276 my $cfs_path = PVE
::QemuConfig-
>cfs_config_path($vmid);
7277 my $oldconf = PVE
::Cluster
::cfs_read_file
($cfs_path);
7278 my $new_conf_raw = '';
7282 my $print_devmap = sub {
7283 my $cfgfn = "$tmpdir/qemu-server.conf";
7285 # we can read the config - that is already extracted
7286 my $fh = IO
::File-
>new($cfgfn, "r") ||
7287 die "unable to read qemu-server.conf - $!\n";
7289 my $fwcfgfn = "$tmpdir/qemu-server.fw";
7291 my $pve_firewall_dir = '/etc/pve/firewall';
7292 mkdir $pve_firewall_dir; # make sure the dir exists
7293 PVE
::Tools
::file_copy
($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
7296 $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
7298 foreach my $info (values %{$virtdev_hash}) {
7299 my $storeid = $info->{storeid
};
7300 next if defined($storage_limits{$storeid});
7302 my $limit = PVE
::Storage
::get_bandwidth_limit
('restore', [$storeid], $bwlimit) // 0;
7303 print STDERR
"rate limit for storage $storeid: $limit KiB/s\n" if $limit;
7304 $storage_limits{$storeid} = $limit * 1024;
7307 foreach my $devname (keys %$devinfo) {
7308 die "found no device mapping information for device '$devname'\n"
7309 if !$devinfo->{$devname}->{virtdev
};
7312 # create empty/temp config
7314 PVE
::Tools
::file_set_contents
($conffile, "memory: 128\n");
7315 $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
7319 my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
7321 # print restore information to $fifofh
7322 foreach my $virtdev (sort keys %$virtdev_hash) {
7323 my $d = $virtdev_hash->{$virtdev};
7324 next if $d->{is_cloudinit
}; # no need to restore cloudinit
7326 my $storeid = $d->{storeid
};
7327 my $volid = $d->{volid
};
7330 if (my $limit = $storage_limits{$storeid}) {
7331 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
7334 my $write_zeros = 1;
7335 if (PVE
::Storage
::volume_has_feature
($cfg, 'sparseinit', $volid)) {
7339 my $path = PVE
::Storage
::path
($cfg, $volid);
7341 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
7343 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
7346 $fh->seek(0, 0) || die "seek failed - $!\n";
7348 my $cookie = { netcount
=> 0 };
7349 while (defined(my $line = <$fh>)) {
7350 $new_conf_raw .= restore_update_config_line
(
7367 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7368 local $SIG{ALRM
} = sub { die "got timeout\n"; };
7370 $oldtimeout = alarm($timeout);
7377 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
7378 my ($dev_id, $size, $devname) = ($1, $2, $3);
7379 $devinfo->{$devname} = { size
=> $size, dev_id
=> $dev_id };
7380 } elsif ($line =~ m/^CTIME: /) {
7381 # we correctly received the vma config, so we can disable
7382 # the timeout now for disk allocation (set to 10 minutes, so
7383 # that we always timeout if something goes wrong)
7386 print $fifofh "done\n";
7387 my $tmp = $oldtimeout || 0;
7388 $oldtimeout = undef;
7395 print "restore vma archive: $dbg_cmdstring\n";
7396 run_command
($commands, input
=> $input, outfunc
=> $parser, afterfork
=> $openfifo);
7400 alarm($oldtimeout) if $oldtimeout;
7402 $restore_deactivate_volumes->($cfg, $virtdev_hash);
7404 close($fifofh) if $fifofh;
7409 $restore_destroy_volumes->($cfg, $virtdev_hash);
7413 my $new_conf = $restore_merge_config->($conffile, $new_conf_raw, $opts->{override_conf
});
7414 PVE
::QemuConfig-
>write_config($vmid, $new_conf);
7416 eval { rescan
($vmid, 1); };
7419 PVE
::AccessControl
::add_vm_to_pool
($vmid, $opts->{pool
}) if $opts->{pool
};
7422 sub restore_tar_archive
{
7423 my ($archive, $vmid, $user, $opts) = @_;
7425 if (scalar(keys $opts->{override_conf
}->%*) > 0) {
7426 my $keystring = join(' ', keys $opts->{override_conf
}->%*);
7427 die "cannot pass along options ($keystring) when restoring from tar archive\n";
7430 if ($archive ne '-') {
7431 my $firstfile = tar_archive_read_firstfile
($archive);
7432 die "ERROR: file '$archive' does not look like a QemuServer vzdump backup\n"
7433 if $firstfile ne 'qemu-server.conf';
7436 my $storecfg = PVE
::Storage
::config
();
7438 # avoid zombie disks when restoring over an existing VM -> cleanup first
7439 # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
7440 # skiplock=1 because qmrestore has set the 'create' lock itself already
7441 my $vmcfgfn = PVE
::QemuConfig-
>config_file($vmid);
7442 destroy_vm
($storecfg, $vmid, 1, { lock => 'restore' }) if -f
$vmcfgfn;
7444 my $tocmd = "/usr/lib/qemu-server/qmextract";
7446 $tocmd .= " --storage " . PVE
::Tools
::shellquote
($opts->{storage
}) if $opts->{storage
};
7447 $tocmd .= " --pool " . PVE
::Tools
::shellquote
($opts->{pool
}) if $opts->{pool
};
7448 $tocmd .= ' --prealloc' if $opts->{prealloc
};
7449 $tocmd .= ' --info' if $opts->{info
};
7451 # tar option "xf" does not autodetect compression when read from STDIN,
7452 # so we pipe to zcat
7453 my $cmd = "zcat -f|tar xf " . PVE
::Tools
::shellquote
($archive) . " " .
7454 PVE
::Tools
::shellquote
("--to-command=$tocmd");
7456 my $tmpdir = "/var/tmp/vzdumptmp$$";
7459 local $ENV{VZDUMP_TMPDIR
} = $tmpdir;
7460 local $ENV{VZDUMP_VMID
} = $vmid;
7461 local $ENV{VZDUMP_USER
} = $user;
7463 my $conffile = PVE
::QemuConfig-
>config_file($vmid);
7464 my $new_conf_raw = '';
7466 # disable interrupts (always do cleanups)
7470 local $SIG{HUP
} = sub { print STDERR
"got interrupt - ignored\n"; };
7478 local $SIG{PIPE
} = sub { die "interrupted by signal\n"; };
7480 if ($archive eq '-') {
7481 print "extracting archive from STDIN\n";
7482 run_command
($cmd, input
=> "<&STDIN");
7484 print "extracting archive '$archive'\n";
7488 return if $opts->{info
};
7492 my $statfile = "$tmpdir/qmrestore.stat";
7493 if (my $fd = IO
::File-
>new($statfile, "r")) {
7494 while (defined (my $line = <$fd>)) {
7495 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
7496 $map->{$1} = $2 if $1;
7498 print STDERR
"unable to parse line in statfile - $line\n";
7504 my $confsrc = "$tmpdir/qemu-server.conf";
7506 my $srcfd = IO
::File-
>new($confsrc, "r") || die "unable to open file '$confsrc'\n";
7508 my $cookie = { netcount
=> 0 };
7509 while (defined (my $line = <$srcfd>)) {
7510 $new_conf_raw .= restore_update_config_line
(
7521 tar_restore_cleanup
($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info
};
7527 PVE
::Tools
::file_set_contents
($conffile, $new_conf_raw);
7529 PVE
::Cluster
::cfs_update
(); # make sure we read new file
7531 eval { rescan
($vmid, 1); };
7535 sub foreach_storage_used_by_vm
{
7536 my ($conf, $func) = @_;
7540 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7541 my ($ds, $drive) = @_;
7542 return if drive_is_cdrom
($drive);
7544 my $volid = $drive->{file
};
7546 my ($sid, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
7547 $sidhash->{$sid} = $sid if $sid;
7550 foreach my $sid (sort keys %$sidhash) {
7555 my $qemu_snap_storage = {
7558 sub do_snapshots_with_qemu
{
7559 my ($storecfg, $volid, $deviceid) = @_;
7561 return if $deviceid =~ m/tpmstate0/;
7563 my $storage_name = PVE
::Storage
::parse_volume_id
($volid);
7564 my $scfg = $storecfg->{ids
}->{$storage_name};
7565 die "could not find storage '$storage_name'\n" if !defined($scfg);
7567 if ($qemu_snap_storage->{$scfg->{type
}} && !$scfg->{krbd
}){
7571 if ($volid =~ m/\.(qcow2|qed)$/){
7578 sub qga_check_running
{
7579 my ($vmid, $nowarn) = @_;
7581 eval { mon_cmd
($vmid, "guest-ping", timeout
=> 3); };
7583 warn "QEMU Guest Agent is not running - $@" if !$nowarn;
7589 sub template_create
{
7590 my ($vmid, $conf, $disk) = @_;
7592 my $storecfg = PVE
::Storage
::config
();
7594 PVE
::QemuConfig-
>foreach_volume($conf, sub {
7595 my ($ds, $drive) = @_;
7597 return if drive_is_cdrom
($drive);
7598 return if $disk && $ds ne $disk;
7600 my $volid = $drive->{file
};
7601 return if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
7603 my $voliddst = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
7604 $drive->{file
} = $voliddst;
7605 $conf->{$ds} = print_drive
($drive);
7606 PVE
::QemuConfig-
>write_config($vmid, $conf);
7610 sub convert_iscsi_path
{
7613 if ($path =~ m
|^iscsi
://([^/]+)/([^/]+)/(.+)$|) {
7618 my $initiator_name = get_initiator_name
();
7620 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
7621 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
7624 die "cannot convert iscsi path '$path', unkown format\n";
7627 sub qemu_img_convert
{
7628 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
7630 my $storecfg = PVE
::Storage
::config
();
7631 my ($src_storeid, $src_volname) = PVE
::Storage
::parse_volume_id
($src_volid, 1);
7632 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid, 1);
7634 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
7638 my $src_is_iscsi = 0;
7642 PVE
::Storage
::activate_volumes
($storecfg, [$src_volid], $snapname);
7643 my $src_scfg = PVE
::Storage
::storage_config
($storecfg, $src_storeid);
7644 $src_format = qemu_img_format
($src_scfg, $src_volname);
7645 $src_path = PVE
::Storage
::path
($storecfg, $src_volid, $snapname);
7646 $src_is_iscsi = ($src_path =~ m
|^iscsi
://|);
7647 $cachemode = 'none' if $src_scfg->{type
} eq 'zfspool';
7648 } elsif (-f
$src_volid || -b
$src_volid) {
7649 $src_path = $src_volid;
7650 if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7655 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
7657 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7658 my $dst_format = qemu_img_format
($dst_scfg, $dst_volname);
7659 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7660 my $dst_is_iscsi = ($dst_path =~ m
|^iscsi
://|);
7663 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
7664 push @$cmd, '-l', "snapshot.name=$snapname"
7665 if $snapname && $src_format && $src_format eq "qcow2";
7666 push @$cmd, '-t', 'none' if $dst_scfg->{type
} eq 'zfspool';
7667 push @$cmd, '-T', $cachemode if defined($cachemode);
7669 if ($src_is_iscsi) {
7670 push @$cmd, '--image-opts';
7671 $src_path = convert_iscsi_path
($src_path);
7672 } elsif ($src_format) {
7673 push @$cmd, '-f', $src_format;
7676 if ($dst_is_iscsi) {
7677 push @$cmd, '--target-image-opts';
7678 $dst_path = convert_iscsi_path
($dst_path);
7680 push @$cmd, '-O', $dst_format;
7683 push @$cmd, $src_path;
7685 if (!$dst_is_iscsi && $is_zero_initialized) {
7686 push @$cmd, "zeroinit:$dst_path";
7688 push @$cmd, $dst_path;
7693 if($line =~ m/\((\S+)\/100\
%\)/){
7695 my $transferred = int($size * $percent / 100);
7696 my $total_h = render_bytes
($size, 1);
7697 my $transferred_h = render_bytes
($transferred, 1);
7699 print "transferred $transferred_h of $total_h ($percent%)\n";
7704 eval { run_command
($cmd, timeout
=> undef, outfunc
=> $parser); };
7706 die "copy failed: $err" if $err;
7709 sub qemu_img_format
{
7710 my ($scfg, $volname) = @_;
7712 if ($scfg->{path
} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
7719 sub qemu_drive_mirror
{
7720 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
7722 $jobs = {} if !$jobs;
7726 $jobs->{"drive-$drive"} = {};
7728 if ($dst_volid =~ /^nbd:/) {
7729 $qemu_target = $dst_volid;
7732 my $storecfg = PVE
::Storage
::config
();
7733 my ($dst_storeid, $dst_volname) = PVE
::Storage
::parse_volume_id
($dst_volid);
7735 my $dst_scfg = PVE
::Storage
::storage_config
($storecfg, $dst_storeid);
7737 $format = qemu_img_format
($dst_scfg, $dst_volname);
7739 my $dst_path = PVE
::Storage
::path
($storecfg, $dst_volid);
7741 $qemu_target = $is_zero_initialized ?
"zeroinit:$dst_path" : $dst_path;
7744 my $opts = { timeout
=> 10, device
=> "drive-$drive", mode
=> "existing", sync
=> "full", target
=> $qemu_target };
7745 $opts->{format
} = $format if $format;
7747 if (defined($src_bitmap)) {
7748 $opts->{sync
} = 'incremental';
7749 $opts->{bitmap
} = $src_bitmap;
7750 print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
7753 if (defined($bwlimit)) {
7754 $opts->{speed
} = $bwlimit * 1024;
7755 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7757 print "drive mirror is starting for drive-$drive\n";
7760 # if a job already runs for this device we get an error, catch it for cleanup
7761 eval { mon_cmd
($vmid, "drive-mirror", %$opts); };
7763 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7765 die "mirroring error: $err\n";
7768 qemu_drive_mirror_monitor
($vmid, $vmiddst, $jobs, $completion, $qga);
7771 # $completion can be either
7772 # 'complete': wait until all jobs are ready, block-job-complete them (default)
7773 # 'cancel': wait until all jobs are ready, block-job-cancel them
7774 # 'skip': wait until all jobs are ready, return with block jobs in ready state
7775 # 'auto': wait until all jobs disappear, only use for jobs which complete automatically
7776 sub qemu_drive_mirror_monitor
{
7777 my ($vmid, $vmiddst, $jobs, $completion, $qga, $op) = @_;
7779 $completion //= 'complete';
7783 my $err_complete = 0;
7785 my $starttime = time ();
7787 die "block job ('$op') timed out\n" if $err_complete > 300;
7789 my $stats = mon_cmd
($vmid, "query-block-jobs");
7792 my $running_jobs = {};
7793 for my $stat (@$stats) {
7794 next if $stat->{type
} ne $op;
7795 $running_jobs->{$stat->{device
}} = $stat;
7798 my $readycounter = 0;
7800 for my $job_id (sort keys %$jobs) {
7801 my $job = $running_jobs->{$job_id};
7803 my $vanished = !defined($job);
7804 my $complete = defined($jobs->{$job_id}->{complete
}) && $vanished;
7805 if($complete || ($vanished && $completion eq 'auto')) {
7806 print "$job_id: $op-job finished\n";
7807 delete $jobs->{$job_id};
7811 die "$job_id: '$op' has been cancelled\n" if !defined($job);
7813 my $busy = $job->{busy
};
7814 my $ready = $job->{ready
};
7815 if (my $total = $job->{len
}) {
7816 my $transferred = $job->{offset
} || 0;
7817 my $remaining = $total - $transferred;
7818 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7820 my $duration = $ctime - $starttime;
7821 my $total_h = render_bytes
($total, 1);
7822 my $transferred_h = render_bytes
($transferred, 1);
7824 my $status = sprintf(
7825 "transferred $transferred_h of $total_h ($percent%%) in %s",
7826 render_duration
($duration),
7831 $status .= ", still busy"; # shouldn't even happen? but mirror is weird
7833 $status .= ", ready";
7836 print "$job_id: $status\n" if !$jobs->{$job_id}->{ready
};
7837 $jobs->{$job_id}->{ready
} = $ready;
7840 $readycounter++ if $job->{ready
};
7843 last if scalar(keys %$jobs) == 0;
7845 if ($readycounter == scalar(keys %$jobs)) {
7846 print "all '$op' jobs are ready\n";
7848 # do the complete later (or has already been done)
7849 last if $completion eq 'skip' || $completion eq 'auto';
7851 if ($vmiddst && $vmiddst != $vmid) {
7852 my $agent_running = $qga && qga_check_running
($vmid);
7853 if ($agent_running) {
7854 print "freeze filesystem\n";
7855 eval { mon_cmd
($vmid, "guest-fsfreeze-freeze"); };
7858 print "suspend vm\n";
7859 eval { PVE
::QemuServer
::vm_suspend
($vmid, 1); };
7863 # if we clone a disk for a new target vm, we don't switch the disk
7864 PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs);
7866 if ($agent_running) {
7867 print "unfreeze filesystem\n";
7868 eval { mon_cmd
($vmid, "guest-fsfreeze-thaw"); };
7871 print "resume vm\n";
7872 eval { PVE
::QemuServer
::vm_resume
($vmid, 1, 1); };
7879 for my $job_id (sort keys %$jobs) {
7880 # try to switch the disk if source and destination are on the same guest
7881 print "$job_id: Completing block job_id...\n";
7884 if ($completion eq 'complete') {
7885 $op = 'block-job-complete';
7886 } elsif ($completion eq 'cancel') {
7887 $op = 'block-job-cancel';
7889 die "invalid completion value: $completion\n";
7891 eval { mon_cmd
($vmid, $op, device
=> $job_id) };
7892 if ($@ =~ m/cannot be completed/) {
7893 print "$job_id: block job cannot be completed, trying again.\n";
7896 print "$job_id: Completed successfully.\n";
7897 $jobs->{$job_id}->{complete
} = 1;
7908 eval { PVE
::QemuServer
::qemu_blockjobs_cancel
($vmid, $jobs) };
7909 die "block job ($op) error: $err";
7913 sub qemu_blockjobs_cancel
{
7914 my ($vmid, $jobs) = @_;
7916 foreach my $job (keys %$jobs) {
7917 print "$job: Cancelling block job\n";
7918 eval { mon_cmd
($vmid, "block-job-cancel", device
=> $job); };
7919 $jobs->{$job}->{cancel
} = 1;
7923 my $stats = mon_cmd
($vmid, "query-block-jobs");
7925 my $running_jobs = {};
7926 foreach my $stat (@$stats) {
7927 $running_jobs->{$stat->{device
}} = $stat;
7930 foreach my $job (keys %$jobs) {
7932 if (defined($jobs->{$job}->{cancel
}) && !defined($running_jobs->{$job})) {
7933 print "$job: Done.\n";
7934 delete $jobs->{$job};
7938 last if scalar(keys %$jobs) == 0;
7945 my ($storecfg, $source, $dest, $full, $newvollist, $jobs, $completion, $qga, $bwlimit) = @_;
7947 my ($vmid, $running) = $source->@{qw(vmid running)};
7948 my ($src_drivename, $drive, $snapname) = $source->@{qw(drivename drive snapname)};
7950 my ($newvmid, $dst_drivename, $efisize) = $dest->@{qw(vmid drivename efisize)};
7951 my ($storage, $format) = $dest->@{qw(storage format)};
7953 my $use_drive_mirror = $full && $running && $src_drivename && !$snapname;
7955 if ($src_drivename && $dst_drivename && $src_drivename ne $dst_drivename) {
7956 die "cloning from/to EFI disk requires EFI disk\n"
7957 if $src_drivename eq 'efidisk0' || $dst_drivename eq 'efidisk0';
7958 die "cloning from/to TPM state requires TPM state\n"
7959 if $src_drivename eq 'tpmstate0' || $dst_drivename eq 'tpmstate0';
7961 # This would lead to two device nodes in QEMU pointing to the same backing image!
7962 die "cannot change drive name when cloning disk from/to the same VM\n"
7963 if $use_drive_mirror && $vmid == $newvmid;
7966 die "cannot move TPM state while VM is running\n"
7967 if $use_drive_mirror && $src_drivename eq 'tpmstate0';
7971 print "create " . ($full ?
'full' : 'linked') . " clone of drive ";
7972 print "$src_drivename " if $src_drivename;
7973 print "($drive->{file})\n";
7976 $newvolid = PVE
::Storage
::vdisk_clone
($storecfg, $drive->{file
}, $newvmid, $snapname);
7977 push @$newvollist, $newvolid;
7980 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($drive->{file
});
7981 $storeid = $storage if $storage;
7983 my $dst_format = resolve_dst_disk_format
($storecfg, $storeid, $volname, $format);
7987 if (drive_is_cloudinit
($drive)) {
7988 $name = "vm-$newvmid-cloudinit";
7989 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
7990 if ($scfg->{path
}) {
7991 $name .= ".$dst_format";
7994 $size = PVE
::QemuServer
::Cloudinit
::CLOUDINIT_DISK_SIZE
;
7995 } elsif ($dst_drivename eq 'efidisk0') {
7996 $size = $efisize or die "internal error - need to specify EFI disk size\n";
7997 } elsif ($dst_drivename eq 'tpmstate0') {
7998 $dst_format = 'raw';
7999 $size = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8001 ($size) = PVE
::Storage
::volume_size_info
($storecfg, $drive->{file
}, 10);
8003 $newvolid = PVE
::Storage
::vdisk_alloc
(
8004 $storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024)
8006 push @$newvollist, $newvolid;
8008 PVE
::Storage
::activate_volumes
($storecfg, [$newvolid]);
8010 if (drive_is_cloudinit
($drive)) {
8011 # when cloning multiple disks (e.g. during clone_vm) it might be the last disk
8012 # if this is the case, we have to complete any block-jobs still there from
8013 # previous drive-mirrors
8014 if (($completion eq 'complete') && (scalar(keys %$jobs) > 0)) {
8015 qemu_drive_mirror_monitor
($vmid, $newvmid, $jobs, $completion, $qga);
8020 my $sparseinit = PVE
::Storage
::volume_has_feature
($storecfg, 'sparseinit', $newvolid);
8021 if ($use_drive_mirror) {
8022 qemu_drive_mirror
($vmid, $src_drivename, $newvolid, $newvmid, $sparseinit, $jobs,
8023 $completion, $qga, $bwlimit);
8025 # TODO: handle bwlimits
8026 if ($dst_drivename eq 'efidisk0') {
8027 # the relevant data on the efidisk may be smaller than the source
8028 # e.g. on RBD/ZFS, so we use dd to copy only the amount
8029 # that is given by the OVMF_VARS.fd
8030 my $src_path = PVE
::Storage
::path
($storecfg, $drive->{file
}, $snapname);
8031 my $dst_path = PVE
::Storage
::path
($storecfg, $newvolid);
8033 my $src_format = (PVE
::Storage
::parse_volname
($storecfg, $drive->{file
}))[6];
8035 # better for Ceph if block size is not too small, see bug #3324
8038 my $cmd = ['qemu-img', 'dd', '-n', '-O', $dst_format];
8040 if ($src_format eq 'qcow2' && $snapname) {
8041 die "cannot clone qcow2 EFI disk snapshot - requires QEMU >= 6.2\n"
8042 if !min_version
(kvm_user_version
(), 6, 2);
8043 push $cmd->@*, '-l', $snapname;
8045 push $cmd->@*, "bs=$bs", "osize=$size", "if=$src_path", "of=$dst_path";
8048 qemu_img_convert
($drive->{file
}, $newvolid, $size, $snapname, $sparseinit);
8054 my ($size) = eval { PVE
::Storage
::volume_size_info
($storecfg, $newvolid, 10) };
8056 my $disk = dclone
($drive);
8057 delete $disk->{format
};
8058 $disk->{file
} = $newvolid;
8059 $disk->{size
} = $size if defined($size);
8064 sub get_running_qemu_version
{
8066 my $res = mon_cmd
($vmid, "query-version");
8067 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
8070 sub qemu_use_old_bios_files
{
8071 my ($machine_type) = @_;
8073 return if !$machine_type;
8075 my $use_old_bios_files = undef;
8077 if ($machine_type =~ m/^(\S+)\.pxe$/) {
8079 $use_old_bios_files = 1;
8081 my $version = extract_version
($machine_type, kvm_user_version
());
8082 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
8083 # load new efi bios files on migration. So this hack is required to allow
8084 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
8085 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
8086 $use_old_bios_files = !min_version
($version, 2, 4);
8089 return ($use_old_bios_files, $machine_type);
8092 sub get_efivars_size
{
8093 my ($conf, $efidisk) = @_;
8095 my $arch = get_vm_arch
($conf);
8096 $efidisk //= $conf->{efidisk0
} ? parse_drive
('efidisk0', $conf->{efidisk0
}) : undef;
8097 my $smm = PVE
::QemuServer
::Machine
::machine_type_is_q35
($conf);
8098 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8099 return -s
$ovmf_vars;
8102 sub update_efidisk_size
{
8105 return if !defined($conf->{efidisk0
});
8107 my $disk = PVE
::QemuServer
::parse_drive
('efidisk0', $conf->{efidisk0
});
8108 $disk->{size
} = get_efivars_size
($conf);
8109 $conf->{efidisk0
} = print_drive
($disk);
8114 sub update_tpmstate_size
{
8117 my $disk = PVE
::QemuServer
::parse_drive
('tpmstate0', $conf->{tpmstate0
});
8118 $disk->{size
} = PVE
::QemuServer
::Drive
::TPMSTATE_DISK_SIZE
;
8119 $conf->{tpmstate0
} = print_drive
($disk);
8122 sub create_efidisk
($$$$$$$) {
8123 my ($storecfg, $storeid, $vmid, $fmt, $arch, $efidisk, $smm) = @_;
8125 my (undef, $ovmf_vars) = get_ovmf_files
($arch, $efidisk, $smm);
8127 my $vars_size_b = -s
$ovmf_vars;
8128 my $vars_size = PVE
::Tools
::convert_size
($vars_size_b, 'b' => 'kb');
8129 my $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
8130 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
8132 qemu_img_convert
($ovmf_vars, $volid, $vars_size_b, undef, 0);
8133 my ($size) = PVE
::Storage
::volume_size_info
($storecfg, $volid, 3);
8135 return ($volid, $size/1024);
8138 sub vm_iothreads_list
{
8141 my $res = mon_cmd
($vmid, 'query-iothreads');
8144 foreach my $iothread (@$res) {
8145 $iothreads->{ $iothread->{id
} } = $iothread->{"thread-id"};
8152 my ($conf, $drive) = @_;
8156 if (!$conf->{scsihw
} || ($conf->{scsihw
} =~ m/^lsi/)) {
8158 } elsif ($conf->{scsihw
} && ($conf->{scsihw
} eq 'virtio-scsi-single')) {
8164 my $controller = int($drive->{index} / $maxdev);
8165 my $controller_prefix = ($conf->{scsihw
} && $conf->{scsihw
} eq 'virtio-scsi-single')
8169 return ($maxdev, $controller, $controller_prefix);
8172 sub resolve_dst_disk_format
{
8173 my ($storecfg, $storeid, $src_volname, $format) = @_;
8174 my ($defFormat, $validFormats) = PVE
::Storage
::storage_default_format
($storecfg, $storeid);
8177 # if no target format is specified, use the source disk format as hint
8179 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8180 $format = qemu_img_format
($scfg, $src_volname);
8186 # test if requested format is supported - else use default
8187 my $supported = grep { $_ eq $format } @$validFormats;
8188 $format = $defFormat if !$supported;
8192 # NOTE: if this logic changes, please update docs & possibly gui logic
8193 sub find_vmstate_storage
{
8194 my ($conf, $storecfg) = @_;
8196 # first, return storage from conf if set
8197 return $conf->{vmstatestorage
} if $conf->{vmstatestorage
};
8199 my ($target, $shared, $local);
8201 foreach_storage_used_by_vm
($conf, sub {
8203 my $scfg = PVE
::Storage
::storage_config
($storecfg, $sid);
8204 my $dst = $scfg->{shared
} ? \
$shared : \
$local;
8205 $$dst = $sid if !$$dst || $scfg->{path
}; # prefer file based storage
8208 # second, use shared storage where VM has at least one disk
8209 # third, use local storage where VM has at least one disk
8210 # fall back to local storage
8211 $target = $shared // $local // 'local';
8217 my ($uuid, $uuid_str);
8218 UUID
::generate
($uuid);
8219 UUID
::unparse
($uuid, $uuid_str);
8223 sub generate_smbios1_uuid
{
8224 return "uuid=".generate_uuid
();
8230 mon_cmd
($vmid, 'nbd-server-stop');
8233 sub create_reboot_request
{
8235 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
8236 or die "failed to create reboot trigger file: $!\n";
8240 sub clear_reboot_request
{
8242 my $path = "/run/qemu-server/$vmid.reboot";
8245 $res = unlink($path);
8246 die "could not remove reboot request for $vmid: $!"
8247 if !$res && $! != POSIX
::ENOENT
;
8252 sub bootorder_from_legacy
{
8253 my ($conf, $bootcfg) = @_;
8255 my $boot = $bootcfg->{legacy
} || $boot_fmt->{legacy
}->{default};
8256 my $bootindex_hash = {};
8258 foreach my $o (split(//, $boot)) {
8259 $bootindex_hash->{$o} = $i*100;
8265 PVE
::QemuConfig-
>foreach_volume($conf, sub {
8266 my ($ds, $drive) = @_;
8268 if (drive_is_cdrom
($drive, 1)) {
8269 if ($bootindex_hash->{d
}) {
8270 $bootorder->{$ds} = $bootindex_hash->{d
};
8271 $bootindex_hash->{d
} += 1;
8273 } elsif ($bootindex_hash->{c
}) {
8274 $bootorder->{$ds} = $bootindex_hash->{c
}
8275 if $conf->{bootdisk
} && $conf->{bootdisk
} eq $ds;
8276 $bootindex_hash->{c
} += 1;
8280 if ($bootindex_hash->{n
}) {
8281 for (my $i = 0; $i < $MAX_NETS; $i++) {
8282 my $netname = "net$i";
8283 next if !$conf->{$netname};
8284 $bootorder->{$netname} = $bootindex_hash->{n
};
8285 $bootindex_hash->{n
} += 1;
8292 # Generate default device list for 'boot: order=' property. Matches legacy
8293 # default boot order, but with explicit device names. This is important, since
8294 # the fallback for when neither 'order' nor the old format is specified relies
8295 # on 'bootorder_from_legacy' above, and it would be confusing if this diverges.
8296 sub get_default_bootdevices
{
8302 my $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 0);
8303 push @ret, $first if $first;
8306 $first = PVE
::QemuServer
::Drive
::resolve_first_disk
($conf, 1);
8307 push @ret, $first if $first;
8310 for (my $i = 0; $i < $MAX_NETS; $i++) {
8311 my $netname = "net$i";
8312 next if !$conf->{$netname};
8313 push @ret, $netname;
8320 sub device_bootorder
{
8323 return bootorder_from_legacy
($conf) if !defined($conf->{boot
});
8325 my $boot = parse_property_string
($boot_fmt, $conf->{boot
});
8328 if (!defined($boot) || $boot->{legacy
}) {
8329 $bootorder = bootorder_from_legacy
($conf, $boot);
8330 } elsif ($boot->{order
}) {
8331 my $i = 100; # start at 100 to allow user to insert devices before us with -args
8332 for my $dev (PVE
::Tools
::split_list
($boot->{order
})) {
8333 $bootorder->{$dev} = $i++;
8340 sub register_qmeventd_handle
{
8344 my $peer = "/var/run/qmeventd.sock";
8349 $fh = IO
::Socket
::UNIX-
>new(Peer
=> $peer, Blocking
=> 0, Timeout
=> 1);
8351 if ($! != EINTR
&& $! != EAGAIN
) {
8352 die "unable to connect to qmeventd socket (vmid: $vmid) - $!\n";
8355 die "unable to connect to qmeventd socket (vmid: $vmid) - timeout "
8356 . "after $count retries\n";
8361 # send handshake to mark VM as backing up
8362 print $fh to_json
({vzdump
=> {vmid
=> "$vmid"}});
8364 # return handle to be closed later when inhibit is no longer required
8368 # bash completion helper
8370 sub complete_backup_archives
{
8371 my ($cmdname, $pname, $cvalue) = @_;
8373 my $cfg = PVE
::Storage
::config
();
8377 if ($cvalue =~ m/^([^:]+):/) {
8381 my $data = PVE
::Storage
::template_list
($cfg, $storeid, 'backup');
8384 foreach my $id (keys %$data) {
8385 foreach my $item (@{$data->{$id}}) {
8386 next if $item->{format
} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
8387 push @$res, $item->{volid
} if defined($item->{volid
});
8394 my $complete_vmid_full = sub {
8397 my $idlist = vmstatus
();
8401 foreach my $id (keys %$idlist) {
8402 my $d = $idlist->{$id};
8403 if (defined($running)) {
8404 next if $d->{template
};
8405 next if $running && $d->{status
} ne 'running';
8406 next if !$running && $d->{status
} eq 'running';
8415 return &$complete_vmid_full();
8418 sub complete_vmid_stopped
{
8419 return &$complete_vmid_full(0);
8422 sub complete_vmid_running
{
8423 return &$complete_vmid_full(1);
8426 sub complete_storage
{
8428 my $cfg = PVE
::Storage
::config
();
8429 my $ids = $cfg->{ids
};
8432 foreach my $sid (keys %$ids) {
8433 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, undef, 1);
8434 next if !$ids->{$sid}->{content
}->{images
};
8441 sub complete_migration_storage
{
8442 my ($cmd, $param, $current_value, $all_args) = @_;
8444 my $targetnode = @$all_args[1];
8446 my $cfg = PVE
::Storage
::config
();
8447 my $ids = $cfg->{ids
};
8450 foreach my $sid (keys %$ids) {
8451 next if !PVE
::Storage
::storage_check_enabled
($cfg, $sid, $targetnode, 1);
8452 next if !$ids->{$sid}->{content
}->{images
};
8461 my $qmpstatus = eval {
8462 PVE
::QemuConfig
::assert_config_exists_on_node
($vmid);
8463 mon_cmd
($vmid, "query-status");
8466 return $qmpstatus && $qmpstatus->{status
} eq "paused";
8469 sub check_volume_storage_type
{
8470 my ($storecfg, $vol) = @_;
8472 my ($storeid, $volname) = PVE
::Storage
::parse_volume_id
($vol);
8473 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storeid);
8474 my ($vtype) = PVE
::Storage
::parse_volname
($storecfg, $vol);
8476 die "storage '$storeid' does not support content-type '$vtype'\n"
8477 if !$scfg->{content
}->{$vtype};
8482 sub add_nets_bridge_fdb
{
8483 my ($conf, $vmid) = @_;
8485 for my $opt (keys %$conf) {
8486 next if $opt !~ m/^net(\d+)$/;
8487 my $iface = "tap${vmid}i$1";
8488 # NOTE: expect setups with learning off to *not* use auto-random-generation of MAC on start
8489 my $net = parse_net
($conf->{$opt}, 1) or next;
8491 my $mac = $net->{macaddr
};
8493 log_warn
("MAC learning disabled, but vNIC '$iface' has no static MAC to add to forwarding DB!")
8494 if !file_read_firstline
("/sys/class/net/$iface/brport/learning");
8498 my $bridge = $net->{bridge
};
8500 PVE
::Network
::SDN
::Zones
::add_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8501 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8502 PVE
::Network
::add_bridge_fdb
($iface, $mac, $net->{firewall
});
8507 sub del_nets_bridge_fdb
{
8508 my ($conf, $vmid) = @_;
8510 for my $opt (keys %$conf) {
8511 next if $opt !~ m/^net(\d+)$/;
8512 my $iface = "tap${vmid}i$1";
8514 my $net = parse_net
($conf->{$opt}) or next;
8515 my $mac = $net->{macaddr
} or next;
8517 my $bridge = $net->{bridge
};
8519 PVE
::Network
::SDN
::Zones
::del_bridge_fdb
($iface, $mac, $bridge, $net->{firewall
});
8520 } elsif (-d
"/sys/class/net/$bridge/bridge") { # avoid fdb management with OVS for now
8521 PVE
::Network
::del_bridge_fdb
($iface, $mac, $net->{firewall
});